library(tidyverse)
## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.0 ──
## ✓ ggplot2 3.3.3 ✓ purrr 0.3.4
## ✓ tibble 3.1.0 ✓ dplyr 1.0.5
## ✓ tidyr 1.1.1 ✓ stringr 1.4.0
## ✓ readr 1.3.1 ✓ forcats 0.4.0
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()
library(here)
## here() starts at /Users/caoanjie/Desktop/projects/SyntacticBootstrappingMA
library(metafor)
## Loading required package: Matrix
##
## Attaching package: 'Matrix'
## The following objects are masked from 'package:tidyr':
##
## expand, pack, unpack
## Loading 'metafor' package (version 2.4-0). For an overview
## and introduction to the package please type: help(metafor).
d <- read_csv(here("data/processed/syntactic_bootstrapping_tidy_data.csv"))
## Parsed with column specification:
## cols(
## .default = col_character(),
## se = col_double(),
## mean_age = col_double(),
## productive_vocab_mean = col_double(),
## productive_vocab_median = col_double(),
## n_train_test_pair = col_double(),
## n_test_trial_per_pair = col_double(),
## n_repetitions_sentence = col_double(),
## n_repetitions_video = col_double(),
## inclusion_certainty = col_double(),
## unique_infant = col_logical(),
## test_method = col_logical(),
## publication_year = col_double(),
## n_1 = col_double(),
## x_1 = col_double(),
## x_2 = col_double(),
## x_2_raw = col_double(),
## sd_1 = col_double(),
## sd_2 = col_double(),
## sd_2_raw = col_double(),
## t = col_double()
## # ... with 5 more columns
## )
## See spec(...) for full column specifications.
alt_d <- read_csv(here("exploratory_analyses/07_PA/alt_ES.csv"))
## Parsed with column specification:
## cols(
## .default = col_character(),
## n_1 = col_double(),
## n_2 = col_double(),
## x_1 = col_double(),
## x_2 = col_double(),
## x_2_raw = col_double(),
## sd_1 = col_double(),
## sd_2 = col_double(),
## sd_2_raw = col_double(),
## t = col_double(),
## d = col_double(),
## d_calc = col_logical()
## )
## See spec(...) for full column specifications.
comparison
alt_id <- alt_d %>%
filter(alternative_calc == "between") %>%
select(unique_id) %>%
pull
ma_subset <- d %>%
filter(unique_id %in% alt_id) %>%
filter(sentence_structure == "transitive") %>%
select(unique_id, short_cite,
same_infant, plot_label, n_1, d_calc, d_var_calc, row_id) %>%
mutate(calc_type = "within") %>%
drop_na()
alt_subset <- alt_d %>%
filter(unique_id %in% alt_id) %>%
rowid_to_column() %>%
mutate(pooled_SD = sqrt(((n_1 - 1) * sd_1 ^ 2 + (n_2 - 1) * sd_2 ^ 2) / (n_1 + n_2 - 2)),
d_calc = (x_1 - x_2) / pooled_SD,
d_var_calc = ((n_1 + n_2) / (n_1 * n_2)) + (d_calc ^ 2 / (2 * (n_1 + n_2)))) %>%
rename(row_id = rowid) %>%
select(unique_id, short_cite,
same_infant, plot_label,n_1, d_calc, d_var_calc, row_id) %>%
mutate(calc_type = "between")%>%
drop_na()
comparison_df <- bind_rows(ma_subset,
alt_subset)
comparison_model <- rma.mv(d_calc ~ calc_type,
V = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id,
method = "REML",
data = comparison_df)
comparison_model
##
## Multivariate Meta-Analysis Model (k = 40; method: REML)
##
## Variance Components:
##
## estim sqrt nlvls fixed factor
## sigma^2.1 0.1443 0.3799 9 no short_cite
## sigma^2.2 0.0000 0.0000 37 no short_cite/same_infant
## sigma^2.3 0.0401 0.2003 39 no short_cite/same_infant/row_id
##
## Test for Residual Heterogeneity:
## QE(df = 38) = 93.4110, p-val < .0001
##
## Test of Moderators (coefficient 2):
## QM(df = 1) = 6.1853, p-val = 0.0129
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## intrcpt 0.8864 0.1697 5.2219 <.0001 0.5537 1.2190 ***
## calc_typewithin -0.3394 0.1365 -2.4870 0.0129 -0.6068 -0.0719 *
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
power analysis
library(metapower)
null_model <- rma(yi = d_calc,
vi = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id,
method = "REML",
data = d )
## Warning: Extra argument ('random') disregarded.
i2 <- null_model$I2
null_model
##
## Random-Effects Model (k = 60; tau^2 estimator: REML)
##
## tau^2 (estimated amount of total heterogeneity): 0.2163 (SE = 0.0573)
## tau (square root of estimated tau^2 value): 0.4651
## I^2 (total heterogeneity / total variability): 72.48%
## H^2 (total variability / sampling variability): 3.63
##
## Test for Heterogeneity:
## Q(df = 59) = 196.0659, p-val < .0001
##
## Model Results:
##
## estimate se zval pval ci.lb ci.ub
## 0.2748 0.0725 3.7913 0.0001 0.1327 0.4169 ***
##
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
i2
## [1] 72.4763
mpower(
effect_size = null_model$beta %>% as.numeric(),
study_size = d %>% select(n_1) %>% pull() %>% mean(),#questionable?
k = nrow(d),
i2 = i2/100,
es_type = "d",
test_type = "two-tailed",
p = 0.05,
con_table = NULL
) %>% plot_mpower()
moderator analysis
transitive_model <- rma(d_calc,
vi = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id,
method = "REML",
data = filter(d, sentence_structure == "transitive") )
## Warning: Extra argument ('random') disregarded.
intransitive_model <- rma(d_calc,
vi = d_var_calc,
random = ~ 1 | short_cite/same_infant/row_id,
method = "REML",
data = filter(d, sentence_structure == "intransitive") )
## Warning: Extra argument ('random') disregarded.
n_study_transitive <- d %>% filter( sentence_structure == "transitive") %>% distinct(plot_label) %>% count() %>% pull()
n_study_intransitive <- d %>% filter( sentence_structure == "intransitive") %>% distinct(plot_label) %>% count() %>% pull()
mod_power(
n_groups = 2,
effect_sizes = c(transitive_model$beta %>% as.numeric(),
intransitive_model$beta %>% as.numeric()),
study_size = n_study_transitive + n_study_intransitive,
k = n_study_transitive,#they are equal
i2 = 0.75,# i2/100, the paper is recommending to use a value like this because this is an average / rule of thumb thing, pg31
es_type = "d",
p = 0.05,
con_table = NULL
) %>% plot_mod_power()
