# Libraries
pacman::p_load(estimatr, texreg, janitor, tidyverse, skimr, compareGroups, BFpack, haven)# Filter coefficients that start with "fac_active"
custom_screen <- function(model) {
coef_names <- names(coef(model))
keep <- grepl("^fac_active", coef_names) # Matches coefficients starting with "fac_active"
coef_list <- coef_names[keep]
setNames(coef_list, coef_list) # Named list where names match values
}df <-
read_csv("data_experimental_sample_v2.csv") %>%
inner_join(read_csv("courseclassify.csv"), by = "course_name") %>%
filter(stem_indi == "STEM") %>%
left_join(read_dta(here::here("data/active_learning_new_constructs_india.dta")), by = "stdid") %>%
mutate(
prior_knowledge_tercile = ntile(prior_knowledge, 3),
stem_self_efficacy_tercile = ntile(stem_self_efficacy, 3),
stem_anxiety_tercile = ntile(stem_anxiety, 3)
)lm1 <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(course_rank ~ fac_active_teaching, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
lm2 <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(course_rank ~ fac_active_teaching + female + ses + reservation_stu + rural + father_college + mother_college, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
lm3 <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(course_rank ~ fac_active_teaching, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm4 <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(course_rank ~ fac_active_teaching + female + ses + reservation_stu + rural + father_college + mother_college + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
lm5 <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1, lm2, lm3, lm4, lm5), custom.note = "%stars", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F)| Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | |
|---|---|---|---|---|---|
| fac_active_teaching | -0.004 | -0.007 | -0.010 | -0.006 | -0.010 |
| (0.013) | (0.013) | (0.013) | (0.013) | (0.013) | |
| female | 0.427*** | 0.427*** | |||
| (0.025) | (0.025) | ||||
| ses | 0.037*** | 0.037*** | |||
| (0.011) | (0.011) | ||||
| reservation_stu | -0.178*** | -0.179*** | |||
| (0.018) | (0.018) | ||||
| rural | -0.076*** | -0.077*** | |||
| (0.013) | (0.013) | ||||
| father_college | 0.019 | 0.019 | |||
| (0.014) | (0.014) | ||||
| mother_college | -0.032** | -0.031** | |||
| (0.014) | (0.014) | ||||
| fac_associate_professor | 0.041 | 0.035 | |||
| (0.036) | (0.034) | ||||
| fac_professor | 0.124* | 0.063 | |||
| (0.065) | (0.057) | ||||
| fac_yearsinhighed | 0.003 | 0.003 | |||
| (0.003) | (0.002) | ||||
| fac_highest_degree_phd | -0.089* | -0.066 | |||
| (0.047) | (0.047) | ||||
| fac_highest_degree_phd_in_prog | -0.021 | -0.055* | |||
| (0.038) | (0.030) | ||||
| fac_degree_college_elite | 0.012 | 0.022 | |||
| (0.025) | (0.021) | ||||
| fac_female | 0.061** | 0.050** | |||
| (0.026) | (0.023) | ||||
| Num. obs. | 33161 | 32248 | 33161 | 32032 | 32942 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | |||||
Model 1:
BF(
x = coef(lm1),
Sigma = vcov(lm1),
n = nobs(lm1),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm1), n = nobs(lm1))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.989 0.007 0.004
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.994
## H2 0.006
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 174.22
## H2 0.006 1.00
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.169 1 29.403 1 174.22 1 4.599752e+75 0.994
## H2 1.000 1 1.000 1 1.00 1 2.718000e+00 0.006
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementModel 2:
BF(
x = coef(lm2),
Sigma = vcov(lm2),
n = nobs(lm2),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm2), n = nobs(lm2))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.987 0.009 0.004
## female 0.000 0.000 1.000
## ses 0.180 0.000 0.820
## reservation_stu 0.000 1.000 0.000
## rural 0.000 1.000 0.000
## father_college 0.971 0.002 0.027
## mother_college 0.878 0.120 0.002
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.994
## H2 0.006
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 157.038
## H2 0.006 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.173 1 27.114 1 157.038 1 1.587803e+68 0.994
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.006
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementModel 3:
BF(
x = coef(lm3),
Sigma = vcov(lm3),
n = nobs(lm3),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm3), n = nobs(lm3))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.985 0.011 0.003
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.993
## H2 0.007
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 135.297
## H2 0.007 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.164 1 22.165 1 135.297 1 5.736206e+58 0.993
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.007
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementModel 4:
BF(
x = coef(lm4),
Sigma = vcov(lm4),
n = nobs(lm4),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm4), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm4), n = nobs(lm4))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.988 0.008 0.004
## female 0.000 0.000 1.000
## ses 0.223 0.000 0.777
## reservation_stu 0.000 1.000 0.000
## rural 0.000 1.000 0.000
## father_college 0.972 0.002 0.026
## mother_college 0.894 0.104 0.002
## fac_associate_professor 0.979 0.003 0.018
## fac_professor 0.934 0.002 0.064
## fac_yearsinhighed 0.976 0.003 0.021
## fac_highest_degree_phd 0.939 0.059 0.002
## fac_highest_degree_phd_in_prog 0.987 0.009 0.004
## fac_degree_college_elite 0.988 0.004 0.008
## fac_female 0.848 0.001 0.150
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.994
## H2 0.006
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 159.691
## H2 0.006 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.171 1 27.387 1 159.691 1 2.253412e+69 0.994
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.006
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementModel 5:
BF(
x = coef(lm5),
Sigma = vcov(lm5),
n = nobs(lm5),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm5), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm5), n = nobs(lm5))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.986 0.011 0.003
## fac_associate_professor 0.982 0.003 0.016
## fac_professor 0.980 0.003 0.017
## fac_yearsinhighed 0.973 0.002 0.025
## fac_highest_degree_phd 0.971 0.027 0.002
## fac_highest_degree_phd_in_prog 0.944 0.054 0.002
## fac_degree_college_elite 0.982 0.003 0.016
## fac_female 0.899 0.002 0.100
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.993
## H2 0.007
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 137.543
## H2 0.007 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.168 1 23.137 1 137.543 1 5.42399e+59 0.993
## H2 1.000 1 1.000 1 1.000 1 2.71800e+00 0.007
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementlm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(female == 0) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(female == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(ses_lowest_tercile == 0) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(ses_lowest_tercile == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(math_lowest_tercile == 0) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(math_lowest_tercile == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm4a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(rural == 0) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm4b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(rural == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm5a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(reservation_stu == 0) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm5b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(reservation_stu == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b, lm4a, lm4b, lm5a, lm5b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Low SES refers to a student being in the bottom tercile of the SES score taken from the first principal component of polychoric principal component analysis on equipment (e.g.~TV, microwave, car, computer) present at a student's home. High score refers to a student being in the top and middle terciles of this SES score. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"), custom.model.names = c("Male", "Female", "High/medium SES", "Low SES", "High math skills", "Low math skills", "Urban", "Rural", "Non-reservation", "Reservation"))| Male | Female | High/medium SES | Low SES | High math skills | Low math skills | Urban | Rural | Non-reservation | Reservation | |
|---|---|---|---|---|---|---|---|---|---|---|
| Active learning | 0.005 | -0.024 | -0.012 | 0.001 | -0.006 | 0.038 | 0.006 | -0.019 | 0.015 | -0.029* |
| (0.014) | (0.017) | (0.015) | (0.017) | (0.018) | (0.025) | (0.014) | (0.017) | (0.016) | (0.016) | |
| Num. obs. | 17680 | 15228 | 20976 | 11514 | 10639 | 4911 | 17379 | 15187 | 15227 | 17671 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Low SES refers to a student being in the bottom tercile of the SES score taken from the first principal component of polychoric principal component analysis on equipment (e.g.~TV, microwave, car, computer) present at a student’s home. High score refers to a student being in the top and middle terciles of this SES score. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||||||||||
Males:
BF(
x = coef(lm1a),
Sigma = vcov(lm1a),
n = nobs(lm1a),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1a), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm1a), n = nobs(lm1a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.984 0.006 0.010
## fac_associate_professor 0.980 0.004 0.016
## fac_professor 0.824 0.002 0.174
## fac_yearsinhighed 0.985 0.006 0.009
## fac_highest_degree_phd 0.983 0.012 0.005
## fac_highest_degree_phd_in_prog 0.985 0.009 0.006
## fac_degree_college_elite 0.970 0.003 0.027
## fac_female 0.982 0.005 0.013
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.992
## H2 0.008
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 125.926
## H2 0.008 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.212 1 26.646 1 125.926 1 4.884092e+54 0.992
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.008
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementFemales:
BF(
x = coef(lm1b),
Sigma = vcov(lm1b),
n = nobs(lm1b),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1b), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm1b), n = nobs(lm1b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.960 0.037 0.003
## fac_associate_professor 0.984 0.007 0.009
## fac_professor 0.984 0.010 0.007
## fac_yearsinhighed 0.880 0.002 0.117
## fac_highest_degree_phd 0.972 0.024 0.004
## fac_highest_degree_phd_in_prog 0.765 0.233 0.002
## fac_degree_college_elite 0.977 0.004 0.019
## fac_female 0.726 0.002 0.272
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.979
## H2 0.021
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 47.776
## H2 0.021 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.189 1 9.02 1 47.776 1 5.608048e+20 0.979
## H2 1.000 1 1.00 1 1.000 1 2.718000e+00 0.021
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementHigh/medium SES:
BF(
x = coef(lm2a),
Sigma = vcov(lm2a),
n = nobs(lm2a),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2a), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm2a), n = nobs(lm2a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.982 0.014 0.004
## fac_associate_professor 0.931 0.002 0.067
## fac_professor 0.977 0.003 0.020
## fac_yearsinhighed 0.974 0.003 0.023
## fac_highest_degree_phd 0.973 0.024 0.003
## fac_highest_degree_phd_in_prog 0.920 0.078 0.002
## fac_degree_college_elite 0.985 0.005 0.010
## fac_female 0.963 0.003 0.035
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.991
## H2 0.009
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 108.266
## H2 0.009 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.178 1 19.316 1 108.266 1 1.045689e+47 0.991
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.009
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementLow SES:
BF(
x = coef(lm2b),
Sigma = vcov(lm2b),
n = nobs(lm2b),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2b), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm2b), n = nobs(lm2b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.982 0.009 0.009
## fac_associate_professor 0.982 0.010 0.008
## fac_professor 0.979 0.006 0.014
## fac_yearsinhighed 0.976 0.005 0.018
## fac_highest_degree_phd 0.974 0.021 0.005
## fac_highest_degree_phd_in_prog 0.975 0.020 0.005
## fac_degree_college_elite 0.978 0.006 0.016
## fac_female 0.883 0.003 0.115
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.991
## H2 0.009
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 107.233
## H2 0.009 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.224 1 23.968 1 107.233 1 3.722303e+46 0.991
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.009
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementHigh math skill:
BF(
x = coef(lm3a),
Sigma = vcov(lm3a),
n = nobs(lm3a),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3a), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm3a), n = nobs(lm3a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.980 0.013 0.007
## fac_associate_professor 0.972 0.005 0.022
## fac_professor 0.972 0.005 0.022
## fac_yearsinhighed 0.977 0.006 0.017
## fac_highest_degree_phd 0.975 0.019 0.006
## fac_highest_degree_phd_in_prog 0.977 0.017 0.006
## fac_degree_college_elite 0.980 0.008 0.012
## fac_female 0.696 0.002 0.302
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.99
## H2 0.01
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.00 97.763
## H2 0.01 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.214 1 20.944 1 97.763 1 2.868917e+42 0.99
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.01
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementLow math skill:
BF(
x = coef(lm3b),
Sigma = vcov(lm3b),
n = nobs(lm3b),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3b), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm3b), n = nobs(lm3b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.920 0.005 0.075
## fac_associate_professor 0.934 0.006 0.060
## fac_professor 0.959 0.008 0.033
## fac_yearsinhighed 0.971 0.011 0.018
## fac_highest_degree_phd 0.856 0.139 0.004
## fac_highest_degree_phd_in_prog 0.010 0.990 0.000
## fac_degree_college_elite 0.857 0.004 0.139
## fac_female 0.923 0.006 0.071
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.958
## H2 0.042
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 22.902
## H2 0.044 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.227 1 5.193 1 22.902 1 8.834554e+09 0.958
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.042
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementUrban:
BF(
x = coef(lm4a),
Sigma = vcov(lm4a),
n = nobs(lm4a),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm4a), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm4a), n = nobs(lm4a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.984 0.005 0.011
## fac_associate_professor 0.985 0.009 0.006
## fac_professor 0.976 0.004 0.020
## fac_yearsinhighed 0.968 0.003 0.028
## fac_highest_degree_phd 0.968 0.028 0.003
## fac_highest_degree_phd_in_prog 0.980 0.016 0.004
## fac_degree_college_elite 0.953 0.003 0.044
## fac_female 0.969 0.003 0.028
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.992
## H2 0.008
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 120.589
## H2 0.008 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.215 1 25.869 1 120.589 1 2.349536e+52 0.992
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.008
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementRural:
BF(
x = coef(lm4b),
Sigma = vcov(lm4b),
n = nobs(lm4b),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm4b), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm4b), n = nobs(lm4b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.970 0.026 0.004
## fac_associate_professor 0.926 0.003 0.072
## fac_professor 0.978 0.005 0.017
## fac_yearsinhighed 0.972 0.004 0.024
## fac_highest_degree_phd 0.970 0.026 0.004
## fac_highest_degree_phd_in_prog 0.891 0.107 0.002
## fac_degree_college_elite 0.984 0.008 0.008
## fac_female 0.742 0.002 0.256
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.985
## H2 0.015
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 65.165
## H2 0.015 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.192 1 12.493 1 65.165 1 1.999432e+28 0.985
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.015
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementNon-reservation:
BF(
x = coef(lm5a),
Sigma = vcov(lm5a),
n = nobs(lm5a),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm5a), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm5a), n = nobs(lm5a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.975 0.004 0.021
## fac_associate_professor 0.982 0.006 0.013
## fac_professor 0.984 0.008 0.008
## fac_yearsinhighed 0.912 0.003 0.086
## fac_highest_degree_phd 0.984 0.010 0.007
## fac_highest_degree_phd_in_prog 0.804 0.194 0.002
## fac_degree_college_elite 0.984 0.007 0.010
## fac_female 0.963 0.003 0.034
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.987
## H2 0.013
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 77.182
## H2 0.013 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.204 1 15.708 1 77.182 1 3.309572e+33 0.987
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.013
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementReservation:
BF(
x = coef(lm5b),
Sigma = vcov(lm5b),
n = nobs(lm5b),
hypothesis = "fac_active_teaching = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm5b), hypothesis = "fac_active_teaching = 0",
## Sigma = vcov(lm5b), n = nobs(lm5b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## fac_active_teaching 0.932 0.066 0.003
## fac_associate_professor 0.981 0.005 0.014
## fac_professor 0.955 0.003 0.042
## fac_yearsinhighed 0.982 0.005 0.013
## fac_highest_degree_phd 0.946 0.051 0.003
## fac_highest_degree_phd_in_prog 0.969 0.028 0.003
## fac_degree_college_elite 0.970 0.003 0.026
## fac_female 0.887 0.002 0.111
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.965
## H2 0.035
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 27.204
## H2 0.037 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.185 1 5.046 1 27.204 1 6.52304e+11 0.965
## H2 1.000 1 1.000 1 1.000 1 2.71800e+00 0.035
##
## Hypotheses:
## H1: fac_active_teaching=0
## H2: complementAll new constructs are z-scored (mean 0, sd 1) at the national sample level.
All variables are aggregated to the student-level. Both models include department fixed-effects.
df_stulevel <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
rename(b_stem_self_efficacy = stem_self_efficacy, b_stem_anxiety = stem_anxiety) %>%
group_by(stdid, department_id, e_stem_self_efficacy, b_stem_self_efficacy, e_stem_anxiety, b_stem_anxiety, female, reservation_stu, father_college, mother_college, ses_lowest_tercile) %>%
summarize(
fac_active_teaching = mean(fac_active_teaching, na.rm = T),
fac_associate_professor = mean(fac_associate_professor, na.rm = T),
fac_professor = mean(fac_professor, na.rm = T),
fac_yearsinhighed = mean(fac_yearsinhighed, na.rm = T),
fac_highest_degree_phd = mean(fac_highest_degree_phd, na.rm = T),
fac_highest_degree_phd_in_prog = mean(fac_highest_degree_phd_in_prog, na.rm = T),
fac_degree_college_elite = mean(fac_degree_college_elite, na.rm = T),
fac_female = mean(fac_female, na.rm = T),
) %>%
ungroup()
lm1 <-
df_stulevel %>%
lm_robust(e_stem_anxiety ~ fac_active_teaching + b_stem_anxiety + female + reservation_stu + father_college + mother_college + ses_lowest_tercile + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ department_id, se_type = "stata")
lm2 <-
df_stulevel %>%
lm_robust(e_stem_self_efficacy ~ fac_active_teaching + b_stem_self_efficacy + female + reservation_stu + father_college + mother_college + ses_lowest_tercile + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ department_id, se_type = "stata")
rm(df_stulevel)
knitreg(list(lm1, lm2), custom.note = "%stars", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.model.names = c("Anxiety", "Self-efficacy"))| Anxiety | Self-efficacy | |
|---|---|---|
| fac_active_teaching | 0.111 | -0.043 |
| (0.131) | (0.124) | |
| b_stem_anxiety | 0.462*** | |
| (0.022) | ||
| female | -0.170*** | 0.116*** |
| (0.049) | (0.044) | |
| reservation_stu | 0.101** | -0.051 |
| (0.042) | (0.039) | |
| father_college | -0.034 | -0.036 |
| (0.048) | (0.044) | |
| mother_college | 0.045 | -0.093** |
| (0.051) | (0.044) | |
| ses_lowest_tercile | 0.023 | -0.009 |
| (0.045) | (0.041) | |
| fac_associate_professor | 0.023 | -0.524* |
| (0.323) | (0.311) | |
| fac_professor | -1.308** | 0.697 |
| (0.566) | (0.526) | |
| fac_yearsinhighed | 0.038 | -0.074*** |
| (0.027) | (0.024) | |
| fac_highest_degree_phd | 0.264 | 0.158 |
| (0.511) | (0.466) | |
| fac_highest_degree_phd_in_prog | 0.020 | -0.112 |
| (0.281) | (0.257) | |
| fac_degree_college_elite | 0.047 | 0.237 |
| (0.339) | (0.283) | |
| fac_female | 0.144 | -0.127 |
| (0.247) | (0.234) | |
| b_stem_self_efficacy | 0.546*** | |
| (0.021) | ||
| Num. obs. | 1873 | 1871 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | ||
In these models, scores above the mean (0) are considered high, scores equal to or below 0 are considered low.
lm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety > mean(stem_anxiety, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety <= mean(stem_anxiety, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy > mean(stem_self_efficacy, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy <= mean(stem_self_efficacy, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge > mean(prior_knowledge, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge <= mean(prior_knowledge, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"), custom.model.names = c("High STEM anxiety", "Low STEM anxiety", "High STEM self-efficacy", "Low STEM self-efficacy", "High prior knowledge", "Low prior knowledge"))| High STEM anxiety | Low STEM anxiety | High STEM self-efficacy | Low STEM self-efficacy | High prior knowledge | Low prior knowledge | |
|---|---|---|---|---|---|---|
| Active learning | -0.022 | 0.007 | -0.006 | -0.012 | -0.017 | 0.001 |
| (0.018) | (0.015) | (0.015) | (0.018) | (0.015) | (0.019) | |
| Num. obs. | 14668 | 15259 | 14667 | 15207 | 13001 | 17040 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||||||
lm1c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(low_anxiety = (stem_anxiety <= mean(stem_anxiety, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * low_anxiety + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
lm2c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_self_efficacy = (stem_self_efficacy > mean(stem_self_efficacy, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * high_self_efficacy + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
lm3c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_prior_knowledge = (prior_knowledge > mean(prior_knowledge, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * high_prior_knowledge + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1c, lm2c, lm3c), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, omit.coef = "(professor)|(highed)|(phd)|(elite)|(female)|(^high)", custom.model.names = c("Anxiety", "Self efficacy", "Prior knowledge"))| Anxiety | Self efficacy | Prior knowledge | |
|---|---|---|---|
| fac_active_teaching | -0.004 | 0.010 | 0.006 |
| (0.019) | (0.018) | (0.018) | |
| low_anxietyTRUE | 0.399*** | ||
| (0.019) | |||
| fac_active_teaching:low_anxietyTRUE | 0.007 | ||
| (0.020) | |||
| fac_active_teaching:high_self_efficacyTRUE | -0.020 | ||
| (0.019) | |||
| fac_active_teaching:high_prior_knowledgeTRUE | -0.016 | ||
| (0.021) | |||
| Num. obs. | 29927 | 29874 | 30041 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | |||
High is defined as the above median, low is defined as the below median.
lm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety > median(stem_anxiety, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety <= median(stem_anxiety, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy > median(stem_self_efficacy, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy <= median(stem_self_efficacy, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge > median(prior_knowledge, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge <= median(prior_knowledge, na.rm = T)) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"), custom.model.names = c("High STEM anxiety", "Low STEM anxiety", "High STEM self-efficacy", "Low STEM self-efficacy", "High prior knowledge", "Low prior knowledge"))| High STEM anxiety | Low STEM anxiety | High STEM self-efficacy | Low STEM self-efficacy | High prior knowledge | Low prior knowledge | |
|---|---|---|---|---|---|---|
| Active learning | -0.022 | 0.007 | -0.006 | -0.012 | -0.010 | -0.002 |
| (0.018) | (0.015) | (0.015) | (0.018) | (0.014) | (0.021) | |
| Num. obs. | 14678 | 15249 | 14886 | 14988 | 14946 | 15095 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||||||
lm1c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(low_anxiety = (stem_anxiety <= median(stem_anxiety, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * low_anxiety + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_self_efficacy = (stem_self_efficacy > median(stem_self_efficacy, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * high_self_efficacy + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_prior_knowledge = (prior_knowledge > median(prior_knowledge, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * high_prior_knowledge + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1c, lm2c, lm3c), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, omit.coef = "(professor)|(highed)|(phd)|(elite)|(female)|(^high)", custom.model.names = c("Anxiety", "Self efficacy", "Prior knowledge"))| Anxiety | Self efficacy | Prior knowledge | |
|---|---|---|---|
| fac_active_teaching | -0.007 | 0.012 | 0.000 |
| (0.016) | (0.017) | (0.017) | |
| low_anxietyTRUE | |||
| fac_active_teaching:low_anxietyTRUE | -0.004 | ||
| (0.016) | |||
| fac_active_teaching:high_self_efficacyTRUE | -0.041*** | ||
| (0.015) | |||
| fac_active_teaching:high_prior_knowledgeTRUE | -0.015 | ||
| (0.014) | |||
| Num. obs. | 29927 | 29874 | 30041 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | |||
High is defined as the highest tercile, low is defined as the lowest tercile.
lm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety_tercile == 3) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_anxiety_tercile == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy_tercile == 3) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(stem_self_efficacy_tercile == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge_tercile == 3) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
filter(prior_knowledge_tercile == 1) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"), custom.model.names = c("High STEM anxiety", "Low STEM anxiety", "High STEM self-efficacy", "Low STEM self-efficacy", "High prior knowledge", "Low prior knowledge"))| High STEM anxiety | Low STEM anxiety | High STEM self-efficacy | Low STEM self-efficacy | High prior knowledge | Low prior knowledge | |
|---|---|---|---|---|---|---|
| Active learning | 0.006 | 0.008 | 0.012 | -0.026 | -0.001 | 0.020 |
| (0.022) | (0.018) | (0.016) | (0.023) | (0.017) | (0.021) | |
| Num. obs. | 9985 | 9945 | 9936 | 9978 | 9996 | 10042 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||||||
lm1c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(low_anxiety = (stem_anxiety_tercile == 1)) %>%
lm_robust(course_rank ~ fac_active_teaching * low_anxiety + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm2c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_self_efficacy = (stem_self_efficacy_tercile == 3)) %>%
lm_robust(course_rank ~ fac_active_teaching * high_self_efficacy + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm3c <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_prior_knowledge = (prior_knowledge_tercile == 3)) %>%
lm_robust(course_rank ~ fac_active_teaching * high_prior_knowledge + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1c, lm2c, lm3c), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, omit.coef = "(professor)|(highed)|(phd)|(elite)|(female)|(^high)|(^low)", custom.model.names = c("Anxiety", "Self efficacy", "Prior knowledge"))| Anxiety | Self efficacy | Prior knowledge | |
|---|---|---|---|
| fac_active_teaching | -0.007 | -0.001 | -0.008 |
| (0.015) | (0.015) | (0.015) | |
| fac_active_teaching:low_anxietyTRUE | -0.005 | ||
| (0.016) | |||
| fac_active_teaching:high_self_efficacyTRUE | -0.021 | ||
| (0.016) | |||
| fac_active_teaching:high_prior_knowledgeTRUE | -0.000 | ||
| (0.014) | |||
| Num. obs. | 29927 | 29874 | 30041 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | |||
# rough tests
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(high_prior_knowledge = (prior_knowledge_tercile == 3)) %>%
select(stdid, contains("prior")) %>%
filter(stdid == "IR045EE1024") %>% view()
distinct(stdid, high_prior_knowledge, prior_knowledge) %>%
filter(stdid == "IR045EE1024")
count(stdid) %>% arrange(-n)Model 1 sample is Low anxiety + High self-efficacy + High prior knowledge.
Model 2 sample is complement of that.
lm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety <= median(stem_anxiety, na.rm = T)) & (stem_self_efficacy > median(stem_self_efficacy, na.rm = T)) & (prior_knowledge > median(prior_knowledge, na.rm = T))) %>%
filter(flag) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety <= median(stem_anxiety, na.rm = T)) & (stem_self_efficacy > median(stem_self_efficacy, na.rm = T)) & (prior_knowledge > median(prior_knowledge, na.rm = T))) %>%
filter(!flag) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"))| Model 1 | Model 2 | |
|---|---|---|
| Active learning | -0.004 | -0.011 |
| (0.020) | (0.015) | |
| Num. obs. | 6137 | 23890 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||
Interaction of this flag [Low anxiety High self-efficacy High prior knowledge] with active learning:
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety <= median(stem_anxiety, na.rm = T)) & (stem_self_efficacy > median(stem_self_efficacy, na.rm = T)) & (prior_knowledge > median(prior_knowledge, na.rm = T))) %>%
lm_robust(course_rank ~ fac_active_teaching * flag + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid) %>%
knitreg(stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, omit.coef = "(professor)|(highed)|(phd)|(elite)|(female)|(^high)")| Model 1 | |
|---|---|
| fac_active_teaching | -0.000 |
| (0.014) | |
| flagTRUE | |
| fac_active_teaching:flagTRUE | -0.035** |
| (0.017) | |
| Num. obs. | 30027 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | |
lm1a <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety_tercile == 1 & stem_self_efficacy_tercile == 3 & prior_knowledge_tercile == 3)) %>%
filter(flag) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
lm1b <-
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety_tercile == 1 & stem_self_efficacy_tercile == 3 & prior_knowledge_tercile == 3)) %>%
filter(!flag) %>%
lm_robust(course_rank ~ fac_active_teaching + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid)
knitreg(list(lm1a, lm1b), custom.note = "Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. ***, **, and * indicate significance at 1%, 5%, and 10%.", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.map = list("fac_active_teaching" = "Active learning"))| Model 1 | Model 2 | |
|---|---|---|
| Active learning | 0.052 | -0.015 |
| (0.034) | (0.014) | |
| Num. obs. | 2642 | 27401 |
| Notes: Each observation is a student-course. The dependent variable is the student course grade on a standardized scale (mean 0, standard deviation 1). The treatment is active learning on a standardized scale (mean 0, standard deviation 1). Grades are provided at the course level and not at the faculty level. All models include faculty controls, student fixed effects, and course fixed effects. Faculty controls include faculty rank, experience, educational qualifications, and gender. Standard errors are clustered at the faculty level. , , and indicate significance at 1%, 5%, and 10%. | ||
Interaction of this flag [Low anxiety High self-efficacy High prior knowledge] with active learning:
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
mutate(flag = (stem_anxiety_tercile == 1 & stem_self_efficacy_tercile == 3 & prior_knowledge_tercile == 3)) %>%
lm_robust(course_rank ~ fac_active_teaching * flag + fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ stdid + fe_var, se_type = "stata", clusters = facid) %>%
knitreg(stars = c(0.01, 0.05, 0.1), include.ci = FALSE, dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, omit.coef = "(professor)|(highed)|(phd)|(elite)|(female)|(^high)")| Model 1 | |
|---|---|
| fac_active_teaching | -0.008 |
| (0.014) | |
| flagTRUE | |
| fac_active_teaching:flagTRUE | -0.001 |
| (0.020) | |
| Num. obs. | 30043 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | |
df %>%
distinct(stdid, .keep_all = T) %>%
skim(female, rural, reservation_stu, father_college, mother_college)| Name | Piped data |
| Number of rows | 2266 |
| Number of columns | 32 |
| _______________________ | |
| Column type frequency: | |
| numeric | 5 |
| ________________________ | |
| Group variables | None |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| female | 2 | 1.00 | 0.44 | 0.50 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▆ |
| rural | 20 | 0.99 | 0.47 | 0.50 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▇ |
| reservation_stu | 4 | 1.00 | 0.54 | 0.50 | 0 | 0 | 1 | 1 | 1 | ▇▁▁▁▇ |
| father_college | 11 | 1.00 | 0.50 | 0.50 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▇ |
| mother_college | 10 | 1.00 | 0.35 | 0.48 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▅ |
Balance:
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ female, data = ., clusters = facid,
## fixed_effects = ~fe_var, se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF
## female -0.0001819 0.00583 -0.0312 0.9751 -0.01164 0.01127 470
##
## Multiple R-squared: 0.694 , Adjusted R-squared: 0.6907
## Multiple R-squared (proj. model): 1.985e-08 , Adjusted R-squared (proj. model): -0.01073
## F-statistic (proj. model): 0.0009737 on 1 and 470 DF, p-value: 0.9751df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ rural, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ rural, data = ., clusters = facid,
## fixed_effects = ~fe_var, se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF
## rural -0.004953 0.005011 -0.9884 0.3235 -0.0148 0.004894 469
##
## Multiple R-squared: 0.6924 , Adjusted R-squared: 0.689
## Multiple R-squared (proj. model): 2.281e-05 , Adjusted R-squared (proj. model): -0.01082
## F-statistic (proj. model): 0.977 on 1 and 469 DF, p-value: 0.3235df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ reservation_stu, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ reservation_stu, data = .,
## clusters = facid, fixed_effects = ~fe_var, se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF
## reservation_stu -0.004702 0.004694 -1.002 0.317 -0.01392 0.004521 470
##
## Multiple R-squared: 0.6937 , Adjusted R-squared: 0.6904
## Multiple R-squared (proj. model): 1.965e-05 , Adjusted R-squared (proj. model): -0.01072
## F-statistic (proj. model): 1.004 on 1 and 470 DF, p-value: 0.317df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ father_college, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ father_college, data = .,
## clusters = facid, fixed_effects = ~fe_var, se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF
## father_college 0.001393 0.005104 0.2729 0.7851 -0.008636 0.01142 470
##
## Multiple R-squared: 0.6942 , Adjusted R-squared: 0.6909
## Multiple R-squared (proj. model): 1.812e-06 , Adjusted R-squared (proj. model): -0.01077
## F-statistic (proj. model): 0.07447 on 1 and 470 DF, p-value: 0.7851df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ mother_college, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ mother_college, data = .,
## clusters = facid, fixed_effects = ~fe_var, se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF
## mother_college -0.002084 0.005104 -0.4082 0.6833 -0.01211 0.007946 470
##
## Multiple R-squared: 0.6941 , Adjusted R-squared: 0.6908
## Multiple R-squared (proj. model): 3.731e-06 , Adjusted R-squared (proj. model): -0.01076
## F-statistic (proj. model): 0.1666 on 1 and 470 DF, p-value: 0.6833df %>%
distinct(facid, .keep_all = T) %>%
skim(fac_associate_professor, fac_professor, fac_yearsinhighed, fac_highest_degree_phd, fac_highest_degree_phd_in_prog, fac_degree_college_elite, fac_female)| Name | Piped data |
| Number of rows | 471 |
| Number of columns | 32 |
| _______________________ | |
| Column type frequency: | |
| numeric | 7 |
| ________________________ | |
| Group variables | None |
Variable type: numeric
| skim_variable | n_missing | complete_rate | mean | sd | p0 | p25 | p50 | p75 | p100 | hist |
|---|---|---|---|---|---|---|---|---|---|---|
| fac_associate_professor | 2 | 1 | 0.17 | 0.38 | 0 | 0 | 0 | 0 | 1 | ▇▁▁▁▂ |
| fac_professor | 2 | 1 | 0.08 | 0.27 | 0 | 0 | 0 | 0 | 1 | ▇▁▁▁▁ |
| fac_yearsinhighed | 2 | 1 | 9.94 | 6.45 | 0 | 6 | 9 | 13 | 46 | ▇▆▁▁▁ |
| fac_highest_degree_phd | 0 | 1 | 0.31 | 0.46 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▃ |
| fac_highest_degree_phd_in_prog | 0 | 1 | 0.15 | 0.36 | 0 | 0 | 0 | 0 | 1 | ▇▁▁▁▂ |
| fac_degree_college_elite | 2 | 1 | 0.31 | 0.46 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▃ |
| fac_female | 1 | 1 | 0.32 | 0.47 | 0 | 0 | 0 | 1 | 1 | ▇▁▁▁▃ |
Balance:
df %>%
mutate(fe_var = str_c(department_id, course_name, sep = "_")) %>%
lm_robust(fac_active_teaching ~ fac_associate_professor + fac_professor + fac_yearsinhighed + fac_highest_degree_phd + fac_highest_degree_phd_in_prog + fac_degree_college_elite + fac_female, data = ., fixed_effects = ~ fe_var, se_type = "stata", clusters = facid) %>%
summary()##
## Call:
## lm_robust(formula = fac_active_teaching ~ fac_associate_professor +
## fac_professor + fac_yearsinhighed + fac_highest_degree_phd +
## fac_highest_degree_phd_in_prog + fac_degree_college_elite +
## fac_female, data = ., clusters = facid, fixed_effects = ~fe_var,
## se_type = "stata")
##
## Standard error type: stata
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|) CI Lower
## fac_associate_professor -0.03009 0.150677 -0.1997 0.84182 -0.326176
## fac_professor 0.20549 0.188621 1.0894 0.27653 -0.165163
## fac_yearsinhighed 0.01091 0.009911 1.1009 0.27149 -0.008564
## fac_highest_degree_phd 0.08272 0.179754 0.4602 0.64558 -0.270503
## fac_highest_degree_phd_in_prog 0.37748 0.163013 2.3157 0.02101 0.057154
## fac_degree_college_elite -0.09287 0.109919 -0.8449 0.39860 -0.308867
## fac_female -0.22270 0.118935 -1.8725 0.06177 -0.456416
## CI Upper DF
## fac_associate_professor 0.26600 467
## fac_professor 0.57614 467
## fac_yearsinhighed 0.03039 467
## fac_highest_degree_phd 0.43595 467
## fac_highest_degree_phd_in_prog 0.69781 467
## fac_degree_college_elite 0.12313 467
## fac_female 0.01101 467
##
## Multiple R-squared: 0.71 , Adjusted R-squared: 0.7069
## Multiple R-squared (proj. model): 0.05246 , Adjusted R-squared (proj. model): 0.04206
## F-statistic (proj. model): 2.172 on 7 and 467 DF, p-value: 0.03547df <- read_csv("data_crosscountry_sample_v3.csv")lm1 <-
df %>%
filter(country == "China") %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm2 <-
df %>%
filter(country == "India") %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm3 <-
df %>%
filter(country == "Russia") %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
knitreg(list(lm1, lm2, lm3), custom.note = "%stars", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, custom.model.names = c("China", "India", "Russia"), dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 3, include.rsquared = F, include.adjrs = F, include.rmse = F, include.nclust = F, custom.coef.names = c("Active learning", "Baseline achievement", "Faculty experience", "Faculty undergrad from elite college", "Faculty graduate from elite college", "Faculty total publications", "Faculty female", "Faculty highest degree is PhD", "Professor", "Associate Professor"), custom.coef.map = list("active_score" = "Active learning"))| China | India | Russia | |
|---|---|---|---|
| Active learning | 0.051 | 0.032 | -0.016 |
| (0.061) | (0.020) | (0.056) | |
| Num. obs. | 1560 | 6618 | 1094 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | |||
Notes:
China:
BF(
x = coef(lm1),
Sigma = vcov(lm1),
n = nobs(lm1),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1), hypothesis = "active_score = 0", Sigma = vcov(lm1),
## n = nobs(lm1))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.933 0.013 0.054
## g1_score 0.949 0.032 0.019
## bfac_yearsinhighed 0.951 0.027 0.022
## bfac_degree_college_elite 0.952 0.025 0.023
## bfac_degree_univ_elite 0.950 0.020 0.029
## bfac_publications 0.928 0.059 0.013
## bfac_female 0.903 0.086 0.011
## bfac_highest_degree_phd 0.950 0.030 0.020
## bfac_professor 0.949 0.018 0.033
## bfac_associate_professor 0.942 0.015 0.043
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.965
## H2 0.035
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 27.649
## H2 0.036 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.166 1 4.582 1 27.649 1 1.017646e+12 0.965
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.035
##
## Hypotheses:
## H1: active_score=0
## H2: complementIndia:
BF(
x = coef(lm2),
Sigma = vcov(lm2),
n = nobs(lm2),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2), hypothesis = "active_score = 0", Sigma = vcov(lm2),
## n = nobs(lm2))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.919 0.004 0.077
## g1_score 0.943 0.052 0.005
## bfac_yearsinhighed 0.975 0.016 0.009
## bfac_degree_college_elite 0.799 0.198 0.003
## bfac_degree_univ_elite 0.965 0.007 0.028
## bfac_publications 0.967 0.026 0.007
## bfac_female 0.971 0.008 0.021
## bfac_highest_degree_phd 0.974 0.009 0.017
## bfac_professor 0.969 0.023 0.007
## bfac_associate_professor 0.976 0.013 0.011
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.958
## H2 0.042
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 22.596
## H2 0.044 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.243 1 5.497 1 22.596 1 6.503363e+09 0.958
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.042
##
## Hypotheses:
## H1: active_score=0
## H2: complementRussia:
BF(
x = coef(lm3),
Sigma = vcov(lm3),
n = nobs(lm3),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3), hypothesis = "active_score = 0", Sigma = vcov(lm3),
## n = nobs(lm3))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.941 0.036 0.023
## g1_score 0.833 0.010 0.157
## bfac_yearsinhighed 0.939 0.040 0.021
## bfac_degree_college_elite NA NA NA
## bfac_degree_univ_elite NA NA NA
## bfac_publications 0.876 0.112 0.012
## bfac_female 0.838 0.152 0.010
## bfac_highest_degree_phd 0.919 0.016 0.066
## bfac_professor 0.939 0.039 0.022
## bfac_associate_professor 0.909 0.077 0.014
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0
## H2 1
##
## Evidence matrix (BFs):
## H1 H2
## H1 1 0
## H2 Inf 1
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 Inf 1 0 1 0 1 1.000 0
## H2 1 1 1 1 1 1 2.718 1
##
## Hypotheses:
## H1: active_score=0
## H2: complementlm1a <-
df %>%
filter(country == "India", female == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm1b <-
df %>%
filter(country == "India", female == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm2a <-
df %>%
filter(country == "China", female == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm2b <-
df %>%
filter(country == "China", female == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm3a <-
df %>%
filter(country == "Russia", female == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm3b <-
df %>%
filter(country == "Russia", female == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b), custom.note = "%stars", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, custom.model.names = c("India Female", "India Male", "China Female", "China Male", "Russia Female", "Russia Male"), dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 2, include.rsquared = F, include.nclust = F, include.adjrs = F, include.rmse = F, custom.coef.names = c("Active learning", "Baseline achievement", "Faculty experience", "Faculty undergrad from elite college", "Faculty graduate from elite college", "Faculty total publications", "Faculty female", "Faculty highest degree is PhD", "Professor", "Associate Professor"), custom.coef.map = list("active_score" = "Active learning"))| India Female | India Male | China Female | China Male | Russia Female | Russia Male | |
|---|---|---|---|---|---|---|
| Active learning | 0.03 | 0.03 | 0.03 | 0.07 | -0.06 | -0.01 |
| (0.03) | (0.02) | (0.14) | (0.08) | (0.08) | (0.06) | |
| Num. obs. | 2594 | 4024 | 452 | 1108 | 238 | 856 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | ||||||
Notes:
India female:
BF(
x = coef(lm1a),
Sigma = vcov(lm1a),
n = nobs(lm1a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1a), hypothesis = "active_score = 0", Sigma = vcov(lm1a),
## n = nobs(lm1a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.943 0.010 0.047
## g1_score 0.962 0.021 0.017
## bfac_yearsinhighed 0.960 0.026 0.014
## bfac_degree_college_elite 0.908 0.084 0.008
## bfac_degree_univ_elite 0.962 0.022 0.017
## bfac_publications 0.956 0.031 0.013
## bfac_female 0.960 0.015 0.025
## bfac_highest_degree_phd 0.954 0.012 0.034
## bfac_professor 0.922 0.070 0.008
## bfac_associate_professor 0.958 0.013 0.029
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.97
## H2 0.03
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.00 32.89
## H2 0.03 1.00
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.237 1 7.809 1 32.89 1 1.923347e+14 0.97
## H2 1.000 1 1.000 1 1.00 1 2.718000e+00 0.03
##
## Hypotheses:
## H1: active_score=0
## H2: complementIndia male:
BF(
x = coef(lm1b),
Sigma = vcov(lm1b),
n = nobs(lm1b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1b), hypothesis = "active_score = 0", Sigma = vcov(lm1b),
## n = nobs(lm1b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.926 0.006 0.068
## g1_score 0.815 0.181 0.004
## bfac_yearsinhighed 0.969 0.018 0.013
## bfac_degree_college_elite 0.716 0.281 0.003
## bfac_degree_univ_elite 0.931 0.007 0.062
## bfac_publications 0.959 0.032 0.009
## bfac_female 0.964 0.010 0.026
## bfac_highest_degree_phd 0.969 0.014 0.017
## bfac_professor 0.968 0.020 0.012
## bfac_associate_professor 0.959 0.032 0.009
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.961
## H2 0.039
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.00 24.971
## H2 0.04 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.26 1 6.493 1 24.971 1 6.993193e+10 0.961
## H2 1.00 1 1.000 1 1.000 1 2.718000e+00 0.039
##
## Hypotheses:
## H1: active_score=0
## H2: complementChina female:
BF(
x = coef(lm2a),
Sigma = vcov(lm2a),
n = nobs(lm2a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2a), hypothesis = "active_score = 0", Sigma = vcov(lm2a),
## n = nobs(lm2a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.912 0.036 0.053
## g1_score 0.543 0.008 0.449
## bfac_yearsinhighed 0.892 0.082 0.026
## bfac_degree_college_elite 0.774 0.015 0.211
## bfac_degree_univ_elite 0.906 0.031 0.063
## bfac_publications 0.832 0.149 0.018
## bfac_female 0.912 0.051 0.037
## bfac_highest_degree_phd 0.913 0.048 0.039
## bfac_professor 0.882 0.024 0.094
## bfac_associate_professor 0.908 0.032 0.059
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.954
## H2 0.046
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 20.624
## H2 0.048 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.138 1 2.852 1 20.624 1 9.055797e+08 0.954
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.046
##
## Hypotheses:
## H1: active_score=0
## H2: complementChina male:
BF(
x = coef(lm2b),
Sigma = vcov(lm2b),
n = nobs(lm2b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2b), hypothesis = "active_score = 0", Sigma = vcov(lm2b),
## n = nobs(lm2b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.917 0.015 0.068
## g1_score 0.734 0.258 0.008
## bfac_yearsinhighed 0.943 0.025 0.032
## bfac_degree_college_elite 0.923 0.061 0.016
## bfac_degree_univ_elite 0.941 0.022 0.037
## bfac_publications 0.938 0.042 0.020
## bfac_female 0.878 0.110 0.012
## bfac_highest_degree_phd 0.942 0.034 0.024
## bfac_professor 0.943 0.026 0.031
## bfac_associate_professor 0.936 0.020 0.045
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.957
## H2 0.043
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 22.055
## H2 0.045 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.159 1 3.517 1 22.055 1 3.789339e+09 0.957
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.043
##
## Hypotheses:
## H1: active_score=0
## H2: complementRussia female:
BF(
x = coef(lm3a),
Sigma = vcov(lm3a),
n = nobs(lm3a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3a), hypothesis = "active_score = 0", Sigma = vcov(lm3a),
## n = nobs(lm3a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.854 0.112 0.033
## g1_score 0.885 0.054 0.061
## bfac_yearsinhighed 0.871 0.090 0.039
## bfac_degree_college_elite NA NA NA
## bfac_degree_univ_elite NA NA NA
## bfac_publications 0.881 0.072 0.046
## bfac_female 0.867 0.095 0.038
## bfac_highest_degree_phd 0.885 0.057 0.058
## bfac_professor 0.885 0.062 0.053
## bfac_associate_professor 0.869 0.092 0.038
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0
## H2 1
##
## Evidence matrix (BFs):
## H1 H2
## H1 1 0
## H2 Inf 1
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 Inf 1 0 1 0 1 1.000 0
## H2 1 1 1 1 1 1 2.718 1
##
## Hypotheses:
## H1: active_score=0
## H2: complementRussia male:
BF(
x = coef(lm3b),
Sigma = vcov(lm3b),
n = nobs(lm3b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3b), hypothesis = "active_score = 0", Sigma = vcov(lm3b),
## n = nobs(lm3b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.936 0.034 0.030
## g1_score 0.799 0.011 0.190
## bfac_yearsinhighed 0.930 0.046 0.023
## bfac_degree_college_elite NA NA NA
## bfac_degree_univ_elite NA NA NA
## bfac_publications 0.868 0.118 0.014
## bfac_female 0.827 0.161 0.012
## bfac_highest_degree_phd 0.908 0.017 0.075
## bfac_professor 0.933 0.041 0.025
## bfac_associate_professor 0.907 0.076 0.017
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0
## H2 1
##
## Evidence matrix (BFs):
## H1 H2
## H1 1 0
## H2 Inf 1
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 Inf 1 0 1 0 1 1.000 0
## H2 1 1 1 1 1 1 2.718 1
##
## Hypotheses:
## H1: active_score=0
## H2: complementlm1a <-
df %>%
filter(country == "India", ses_lowest_tercile == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm1b <-
df %>%
filter(country == "India", ses_lowest_tercile == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm2a <-
df %>%
filter(country == "China", ses_lowest_tercile == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm2b <-
df %>%
filter(country == "China", ses_lowest_tercile == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm3a <-
df %>%
filter(country == "Russia", ses_lowest_tercile == 0) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
lm3b <-
df %>%
filter(country == "Russia", ses_lowest_tercile == 1) %>%
lm_robust(g3_score ~ active_score + g1_score + bfac_yearsinhighed + bfac_degree_college_elite + bfac_degree_univ_elite + bfac_publications + bfac_female + bfac_highest_degree_phd + bfac_professor + bfac_associate_professor, fixed_effects = ~ stdid, data = ., clusters = department_id, se_type = "stata")
knitreg(list(lm1a, lm1b, lm2a, lm2b, lm3a, lm3b), custom.note = "%stars", stars = c(0.01, 0.05, 0.1), include.ci = FALSE, custom.model.names = c("India High SES", "India Low SES", "China High SES", "China Low SES", "Russia High SES", "Russia Low SES"), dcolumn = TRUE, booktabs = TRUE, float.pos = "H", caption = "", digits = 2, include.rsquared = F, include.adjrs = F, include.nclust = F, include.rmse = F, custom.coef.names = c("Active learning", "Baseline achievement", "Faculty experience", "Faculty undergrad from elite college", "Faculty graduate from elite college", "Faculty total publications", "Faculty female", "Faculty highest degree is PhD", "Professor", "Associate Professor"), custom.coef.map = list("active_score" = "Active learning"))| India High SES | India Low SES | China High SES | China Low SES | Russia High SES | Russia Low SES | |
|---|---|---|---|---|---|---|
| Active learning | 0.04 | 0.01 | -0.01 | 0.16** | -0.00 | -0.04 |
| (0.02) | (0.04) | (0.09) | (0.07) | (0.07) | (0.09) | |
| Num. obs. | 4490 | 2080 | 944 | 616 | 704 | 390 |
| ***p < 0.01; **p < 0.05; *p < 0.1 | ||||||
Notes:
India high SES:
BF(
x = coef(lm1a),
Sigma = vcov(lm1a),
n = nobs(lm1a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1a), hypothesis = "active_score = 0", Sigma = vcov(lm1a),
## n = nobs(lm1a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.896 0.005 0.099
## g1_score 0.927 0.067 0.006
## bfac_yearsinhighed 0.969 0.020 0.011
## bfac_degree_college_elite 0.769 0.227 0.004
## bfac_degree_univ_elite 0.964 0.009 0.027
## bfac_publications 0.963 0.028 0.009
## bfac_female 0.957 0.008 0.035
## bfac_highest_degree_phd 0.968 0.010 0.022
## bfac_professor 0.965 0.026 0.009
## bfac_associate_professor 0.971 0.015 0.014
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.945
## H2 0.055
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 17.214
## H2 0.058 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.257 1 4.432 1 17.214 1 29915761.375 0.945
## H2 1.000 1 1.000 1 1.000 1 2.718 0.055
##
## Hypotheses:
## H1: active_score=0
## H2: complementIndia low SES:
BF(
x = coef(lm1b),
Sigma = vcov(lm1b),
n = nobs(lm1b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm1b), hypothesis = "active_score = 0", Sigma = vcov(lm1b),
## n = nobs(lm1b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.956 0.016 0.028
## g1_score 0.948 0.039 0.013
## bfac_yearsinhighed 0.957 0.026 0.017
## bfac_degree_college_elite 0.942 0.046 0.012
## bfac_degree_univ_elite 0.935 0.011 0.054
## bfac_publications 0.949 0.038 0.013
## bfac_female 0.954 0.031 0.015
## bfac_highest_degree_phd 0.958 0.023 0.020
## bfac_professor 0.957 0.026 0.017
## bfac_associate_professor 0.958 0.020 0.022
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.977
## H2 0.023
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 43.246
## H2 0.023 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.218 1 9.434 1 43.246 1 6.04472e+18 0.977
## H2 1.000 1 1.000 1 1.000 1 2.71800e+00 0.023
##
## Hypotheses:
## H1: active_score=0
## H2: complementChina high SES:
BF(
x = coef(lm2a),
Sigma = vcov(lm2a),
n = nobs(lm2a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2a), hypothesis = "active_score = 0", Sigma = vcov(lm2a),
## n = nobs(lm2a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.939 0.033 0.029
## g1_score 0.939 0.031 0.030
## bfac_yearsinhighed 0.929 0.020 0.051
## bfac_degree_college_elite 0.938 0.035 0.027
## bfac_degree_univ_elite 0.936 0.025 0.039
## bfac_publications 0.927 0.053 0.020
## bfac_female 0.913 0.070 0.017
## bfac_highest_degree_phd 0.937 0.025 0.038
## bfac_professor 0.939 0.030 0.031
## bfac_associate_professor 0.936 0.024 0.040
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.968
## H2 0.032
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 30.627
## H2 0.033 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.148 1 4.519 1 30.627 1 2.000742e+13 0.968
## H2 1.000 1 1.000 1 1.000 1 2.718000e+00 0.032
##
## Hypotheses:
## H1: active_score=0
## H2: complementChina low SES:
BF(
x = coef(lm2b),
Sigma = vcov(lm2b),
n = nobs(lm2b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm2b), hypothesis = "active_score = 0", Sigma = vcov(lm2b),
## n = nobs(lm2b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.514 0.006 0.480
## g1_score 0.901 0.078 0.021
## bfac_yearsinhighed 0.858 0.126 0.016
## bfac_degree_college_elite 0.922 0.029 0.049
## bfac_degree_univ_elite 0.925 0.036 0.039
## bfac_publications 0.894 0.086 0.020
## bfac_female 0.890 0.090 0.020
## bfac_highest_degree_phd 0.886 0.095 0.019
## bfac_professor 0.921 0.028 0.051
## bfac_associate_professor 0.910 0.023 0.067
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0.679
## H2 0.321
##
## Evidence matrix (BFs):
## H1 H2
## H1 1.000 2.111
## H2 0.474 1.000
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 0.228 1 0.481 1 2.111 1 8.258 0.679
## H2 1.000 1 1.000 1 1.000 1 2.718 0.321
##
## Hypotheses:
## H1: active_score=0
## H2: complementRussia high SES:
BF(
x = coef(lm3a),
Sigma = vcov(lm3a),
n = nobs(lm3a),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3a), hypothesis = "active_score = 0", Sigma = vcov(lm3a),
## n = nobs(lm3a))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.930 0.035 0.035
## g1_score 0.853 0.015 0.132
## bfac_yearsinhighed 0.923 0.051 0.025
## bfac_degree_college_elite NA NA NA
## bfac_degree_univ_elite NA NA NA
## bfac_publications 0.837 0.150 0.014
## bfac_female 0.791 0.197 0.012
## bfac_highest_degree_phd 0.915 0.022 0.063
## bfac_professor 0.923 0.051 0.025
## bfac_associate_professor 0.902 0.078 0.019
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0
## H2 1
##
## Evidence matrix (BFs):
## H1 H2
## H1 1 0
## H2 Inf 1
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 Inf 1 0 1 0 1 1.000 0
## H2 1 1 1 1 1 1 2.718 1
##
## Hypotheses:
## H1: active_score=0
## H2: complementRussia low SES:
BF(
x = coef(lm3b),
Sigma = vcov(lm3b),
n = nobs(lm3b),
hypothesis = "active_score = 0"
) %>%
summary()## Call:
## BF.default(x = coef(lm3b), hypothesis = "active_score = 0", Sigma = vcov(lm3b),
## n = nobs(lm3b))
##
## Bayesian hypothesis test
## Type: exploratory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities
##
## Model parameters:
## Pr(=0) Pr(<0) Pr(>0)
## active_score 0.900 0.066 0.034
## g1_score 0.669 0.012 0.319
## bfac_yearsinhighed 0.907 0.052 0.041
## bfac_degree_college_elite NA NA NA
## bfac_degree_univ_elite NA NA NA
## bfac_publications 0.870 0.105 0.025
## bfac_female 0.846 0.133 0.021
## bfac_highest_degree_phd 0.875 0.025 0.100
## bfac_professor 0.908 0.044 0.048
## bfac_associate_professor 0.875 0.100 0.025
##
## Bayesian hypothesis test
## Type: confirmatory
## Object: numeric
## Parameter: general parameters
## Method: adjusted fractional Bayes factors using Gaussian approximations
##
## Posterior probabilities:
## Pr(hypothesis|data)
## H1 0
## H2 1
##
## Evidence matrix (BFs):
## H1 H2
## H1 1 0
## H2 Inf 1
##
## Specification table:
## complex= complex> fit= fit> BF= BF> BF PHP
## H1 Inf 1 0 1 0 1 1.000 0
## H2 1 1 1 1 1 1 2.718 1
##
## Hypotheses:
## H1: active_score=0
## H2: complementdf %>%
distinct(stdid, .keep_all = T) %>%
compareGroups(country ~ female + age_years + father_college + mother_college, data = ., include.label = F) %>%
createTable(show.all = F, show.n = F, show.p.overall = F)——–Summary descriptives table by ‘country’———
China India Russia
N=780 N=3309 N=547 ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ female 0.29 (0.45) 0.39 (0.49) 0.22 (0.41) age_years 21.3 (1.10) 20.2 (0.91) 20.5 (1.30) father_college 0.20 (0.40) 0.59 (0.49) 0.82 (0.38) mother_college 0.15 (0.35) 0.45 (0.50) 0.89 (0.32) ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
df %>%
distinct(stdid, .keep_all = T) %>%
filter(country == "India") %>%
inner_join(read_csv(here::here("data/depweights.csv")), by = "department_id") %>%
summarise(x = weighted.mean(father_college, w = sw_f, na.rm = T)) x