load libraries.

library(psych)
library(lavaan)
## This is lavaan 0.6-14
## lavaan is FREE software! Please report any bugs.
## 
## Attaching package: 'lavaan'
## The following object is masked from 'package:psych':
## 
##     cor2cov
library(semTools)
## 
## ###############################################################################
## This is semTools 0.5-6
## All users of R (or SEM) are invited to submit functions or ideas for functions.
## ###############################################################################
## 
## Attaching package: 'semTools'
## The following objects are masked from 'package:psych':
## 
##     reliability, skew
library(semPlot)
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyverse)
## ── Attaching packages
## ───────────────────────────────────────
## tidyverse 1.3.2 ──
## ✔ ggplot2 3.4.0     ✔ purrr   0.3.4
## ✔ tibble  3.1.8     ✔ stringr 1.4.1
## ✔ tidyr   1.2.1     ✔ forcats 0.5.2
## ✔ readr   2.1.2     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ ggplot2::%+%()     masks psych::%+%()
## ✖ ggplot2::alpha()   masks psych::alpha()
## ✖ readr::clipboard() masks semTools::clipboard()
## ✖ dplyr::filter()    masks stats::filter()
## ✖ dplyr::lag()       masks stats::lag()

load data.

data <- read.csv("~/okara004@ucr.edu - Google Drive/My Drive/classes/3 -- spring '23/PSYC 214 -- stats/homework/HW #1/HW1_Data/WLS_HW1_CFA_rev23.csv")

get descriptives.

# gender
gender <- factor(data$sexrsp, levels = 1:2, labels = c("Male", "Female"))
summary(gender)
##   Male Female 
##   4637   5051
4637 + 5051 ### = 9688
## [1] 9688
5051/9688 ### = 0.5213666, 52.14% female
## [1] 0.5213666
# age
data <- mutate(data, Age = 103-(data$byear)) 
psych::describe(data$Age) ### mean age: 64.16yrs; SD: 0.51yrs
##    vars    n  mean   sd median trimmed mad min max range skew kurtosis   se
## X1    1 9688 64.16 0.51     64    64.1   0  63  66     3  1.1     2.71 0.01
# education
data$educ92 <- as.numeric(data$educ92)
psych::describe(data$educ92) ### mean: 13.61 (about 1yr of college); SD: 2.26yrs
##    vars    n  mean   sd median trimmed mad min max range skew kurtosis   se
## X1    1 8452 13.61 2.26     12   13.21   0  12  21     9 1.14     0.04 0.02
# nationality on father's side
glimpse(data$ge096fa)
##  int [1:9688] NA 14 4 NA 2 14 14 NA 14 14 ...
data$ge096fa <- as.factor(data$ge096fa)
summary(data$ge096fa) ### primarily european
##    1    2    3    4    5    6    7    8    9   10   11   12   13   14   15   16 
##  633  108   52  463  549  166  128  182    6   82   81  231    4 3417  474  202 
##   17   18   19   20   22   23   24   25   26   27   28   29   30   31   32   33 
##   43   21   47    4   20   10   91    3    8   21   55    2    1    1    1    1 
##   34   35   36   37   38   39   40   41   42 NA's 
##    1    5   16   11   20    9   16   11    4 2488
# nationality on mother's side
glimpse(data$ge095ma)
##  int [1:9688] NA 4 1 NA 41 14 4 NA 14 14 ...
data$ge095ma <- as.factor(data$ge095ma)
summary(data$ge095ma) ### also primarily european
##    1    2    3    4    5    6    7    8    9   10   11   12   13   14   15   16 
##  622   91   38  495  554  151  104  168   28   75   81  218   14 3106  527  192 
##   17   18   19   20   22   23   24   25   26   27   28   30   32   33   35   36 
##   48   31   59    3   50    7   80    5    8   32   36    1    1    1    2   16 
##   37   38   39   40   41   42   43   44   46   47 NA's 
##    8   58   26   27   32   11    3    1    1    1 2676

observed corr & cov/var matrices.

# take a look at the observed correlation matrix.
cor_matrix <- cor(data[, c("in501rer", "in502rer", "in503rer", "in504rer", "in505rer", "in506rer", "in507rer")], use = "complete", method = "pearson")
print(cor_matrix)
##           in501rer  in502rer  in503rer  in504rer  in505rer  in506rer  in507rer
## in501rer 1.0000000 0.2792627 0.3178501 0.3187617 0.3245953 0.2274303 0.2614723
## in502rer 0.2792627 1.0000000 0.3957867 0.4068193 0.3168857 0.2216130 0.2816966
## in503rer 0.3178501 0.3957867 1.0000000 0.3038079 0.2239655 0.1543646 0.2044126
## in504rer 0.3187617 0.4068193 0.3038079 1.0000000 0.6495276 0.3410393 0.3231119
## in505rer 0.3245953 0.3168857 0.2239655 0.6495276 1.0000000 0.4479959 0.3972244
## in506rer 0.2274303 0.2216130 0.1543646 0.3410393 0.4479959 1.0000000 0.5039401
## in507rer 0.2614723 0.2816966 0.2044126 0.3231119 0.3972244 0.5039401 1.0000000
### correlation between volunteer identity and group/organization identity is pretty high compared to the others. correlations among factor 1 variables are not all that much higher than the correlations between factor 1 variables and factor 2 variables, but correlations among the factor 2 variables do appear to be a bit higher than correlations between the factor 2 and factor 1 variables.

# covariance matrix
cov_matrix <- cov(data[, c("in501rer", "in502rer", "in503rer", "in504rer", "in505rer", "in506rer", "in507rer")], use = "complete")
print(cov_matrix)
##           in501rer  in502rer  in503rer  in504rer  in505rer  in506rer  in507rer
## in501rer 3.4519497 1.0836549 0.8668438 1.0220569 1.0973510 0.7619236 0.9173476
## in502rer 1.0836549 4.3620590 1.2133700 1.4663040 1.2042580 0.8345877 1.1109728
## in503rer 0.8668438 1.2133700 2.1546295 0.7695947 0.5981891 0.4085685 0.5665912
## in504rer 1.0220569 1.4663040 0.7695947 2.9781975 2.0396021 1.0612363 1.0529450
## in505rer 1.0973510 1.2042580 0.5981891 2.0396021 3.3108706 1.4698602 1.3648441
## in506rer 0.7619236 0.8345877 0.4085685 1.0612363 1.4698602 3.2513359 1.7158757
## in507rer 0.9173476 1.1109728 0.5665912 1.0529450 1.3648441 1.7158757 3.5657587

two-factor model.

# start by fitting a two-factor model.
model_2f <- "
  F1 = ~ in501rer + in502rer + in503rer
  F2 = ~ in504rer + in505rer + in506rer + in507rer"

# run the CFA and take a look at the fit indices.
model_2f_fit <- cfa(model_2f, data = data)
summary(model_2f_fit, standard = TRUE, fit.measures = TRUE) 
## lavaan 0.6.14 ended normally after 37 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        15
## 
##                                                   Used       Total
##   Number of observations                          6312        9688
## 
## Model Test User Model:
##                                                       
##   Test statistic                              1283.951
##   Degrees of freedom                                13
##   P-value (Chi-square)                           0.000
## 
## Model Test Baseline Model:
## 
##   Test statistic                             11105.204
##   Degrees of freedom                                21
##   P-value                                        0.000
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.885
##   Tucker-Lewis Index (TLI)                       0.815
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)             -83717.758
##   Loglikelihood unrestricted model (H1)     -83075.783
##                                                       
##   Akaike (AIC)                              167465.517
##   Bayesian (BIC)                            167566.770
##   Sample-size adjusted Bayesian (SABIC)     167519.104
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.124
##   90 Percent confidence interval - lower         0.119
##   90 Percent confidence interval - upper         0.130
##   P-value H_0: RMSEA <= 0.050                    0.000
##   P-value H_0: RMSEA >= 0.080                    1.000
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.058
## 
## Parameter Estimates:
## 
##   Standard errors                             Standard
##   Information                                 Expected
##   Information saturated (h1) model          Structured
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 =~                                                                 
##     in501rer          1.000                               0.984    0.530
##     in502rer          1.352    0.047   28.747    0.000    1.330    0.637
##     in503rer          0.839    0.030   27.522    0.000    0.825    0.562
##   F2 =~                                                                 
##     in504rer          1.000                               1.315    0.762
##     in505rer          1.122    0.021   52.958    0.000    1.475    0.811
##     in506rer          0.754    0.019   39.452    0.000    0.991    0.550
##     in507rer          0.755    0.020   37.750    0.000    0.992    0.526
## 
## Covariances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 ~~                                                                 
##     F2                0.896    0.035   25.726    0.000    0.693    0.693
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .in501rer          2.484    0.054   45.658    0.000    2.484    0.720
##    .in502rer          2.594    0.069   37.437    0.000    2.594    0.595
##    .in503rer          1.474    0.034   43.616    0.000    1.474    0.684
##    .in504rer          1.249    0.034   36.875    0.000    1.249    0.419
##    .in505rer          1.134    0.037   30.436    0.000    1.134    0.342
##    .in506rer          2.269    0.045   50.640    0.000    2.269    0.698
##    .in507rer          2.580    0.050   51.314    0.000    2.580    0.724
##     F1                0.968    0.053   18.210    0.000    1.000    1.000
##     F2                1.729    0.054   31.734    0.000    1.000    1.000
### none of the fit indices are looking too good...chi-square is very large for only 13 degrees of freedom, CFI & TLI do not reach the 0.95 threshold, and the RMSEA does not fall below the .06 threshold

# visualization of the model.
semPaths(model_2f_fit)

# take a look at the expected cov matrix and the residual cov matrix.
exp_matrix_2f <- fitted(model_2f_fit)
exp_matrix_2f
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  3.451                                          
## in502rer  1.308  4.361                                   
## in503rer  0.812  1.097  2.154                            
## in504rer  0.896  1.211  0.751  2.978                     
## in505rer  1.005  1.359  0.843  1.940  3.310              
## in506rer  0.675  0.913  0.566  1.303  1.462  3.251       
## in507rer  0.676  0.914  0.567  1.305  1.464  0.984  3.565
resid_matrix_2f <- resid(model_2f_fit)
resid_matrix_2f
## $type
## [1] "raw"
## 
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  0.000                                          
## in502rer -0.225  0.000                                   
## in503rer  0.055  0.116  0.000                            
## in504rer  0.126  0.255  0.018  0.000                     
## in505rer  0.092 -0.155 -0.245  0.099  0.000              
## in506rer  0.086 -0.078 -0.158 -0.242  0.007  0.000       
## in507rer  0.241  0.197 -0.001 -0.252 -0.100  0.732  0.000

one-factor model.

# let's see if the one-factor model fits better than the two-factor model.
model_1f <- "
  F1 = ~ in501rer + in502rer + in503rer + in504rer + in505rer + in506rer + in507rer"

# run the CFA and take a look at the fit indices.
model_1f_fit <- cfa(model_1f, data = data)
summary(model_1f_fit, standard = TRUE, fit.measures = TRUE)
## lavaan 0.6.14 ended normally after 33 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        14
## 
##                                                   Used       Total
##   Number of observations                          6312        9688
## 
## Model Test User Model:
##                                                       
##   Test statistic                              1858.735
##   Degrees of freedom                                14
##   P-value (Chi-square)                           0.000
## 
## Model Test Baseline Model:
## 
##   Test statistic                             11105.204
##   Degrees of freedom                                21
##   P-value                                        0.000
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.834
##   Tucker-Lewis Index (TLI)                       0.750
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)             -84005.150
##   Loglikelihood unrestricted model (H1)     -83075.783
##                                                       
##   Akaike (AIC)                              168038.301
##   Bayesian (BIC)                            168132.804
##   Sample-size adjusted Bayesian (SABIC)     168088.315
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.144
##   90 Percent confidence interval - lower         0.139
##   90 Percent confidence interval - upper         0.150
##   P-value H_0: RMSEA <= 0.050                    0.000
##   P-value H_0: RMSEA >= 0.080                    1.000
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.072
## 
## Parameter Estimates:
## 
##   Standard errors                             Standard
##   Information                                 Expected
##   Information saturated (h1) model          Structured
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 =~                                                                 
##     in501rer          1.000                               0.840    0.452
##     in502rer          1.241    0.046   26.990    0.000    1.042    0.499
##     in503rer          0.689    0.030   23.317    0.000    0.579    0.394
##     in504rer          1.563    0.048   32.499    0.000    1.312    0.760
##     in505rer          1.693    0.052   32.716    0.000    1.421    0.781
##     in506rer          1.170    0.041   28.303    0.000    0.982    0.545
##     in507rer          1.198    0.043   27.973    0.000    1.006    0.533
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .in501rer          2.746    0.052   52.954    0.000    2.746    0.796
##    .in502rer          3.276    0.063   51.999    0.000    3.276    0.751
##    .in503rer          1.820    0.034   53.875    0.000    1.820    0.845
##    .in504rer          1.256    0.033   38.064    0.000    1.256    0.422
##    .in505rer          1.291    0.036   35.663    0.000    1.291    0.390
##    .in506rer          2.286    0.045   50.833    0.000    2.286    0.703
##    .in507rer          2.554    0.050   51.167    0.000    2.554    0.716
##     F1                0.705    0.041   17.178    0.000    1.000    1.000
### chi-squared is larger than the two-factor model, CFI and TLI are both smaller than the two-factor model, AIC and BIC are both larger than the two-factor model, and RMSEA is larger than the two-factor model. so we gained one degree of freedom, but fit worsened.

# visualization of the model.
semPaths(model_1f_fit)

# take a look at the expected cov matrix and the residual cov matrix.
exp_matrix_1f <- fitted(model_1f_fit)
exp_matrix_1f
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  3.451                                          
## in502rer  0.875  4.361                                   
## in503rer  0.486  0.603  2.154                            
## in504rer  1.102  1.367  0.759  2.978                     
## in505rer  1.193  1.481  0.822  1.865  3.310              
## in506rer  0.825  1.023  0.568  1.289  1.396  3.251       
## in507rer  0.845  1.048  0.582  1.320  1.429  0.988  3.565
resid_matrix_1f <- resid(model_1f_fit)
resid_matrix_1f
## $type
## [1] "raw"
## 
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  0.000                                          
## in502rer  0.209  0.000                                   
## in503rer  0.381  0.610  0.000                            
## in504rer -0.080  0.099  0.010  0.000                     
## in505rer -0.096 -0.277 -0.224  0.175  0.000              
## in506rer -0.063 -0.189 -0.160 -0.228  0.074  0.000       
## in507rer  0.073  0.063 -0.015 -0.267 -0.065  0.728  0.000

two-factor model – vol & org equated.

# since volunteer identity and group/organization identity had the highest correlation on in the observed correlation matrix, i'm going to try equating them in the model so see if that improves fit.
model_2fEQ <- "
  F1 = ~ in501rer + in502rer + in503rer
  F2 = ~ a*in504rer + a*in505rer + in506rer + in507rer"

# run the CFA and take a look at the fit indices.
model_2fEQ_fit <- cfa(model_2fEQ, data = data)
summary(model_2fEQ_fit, standard = TRUE, fit.measures = TRUE)
## lavaan 0.6.14 ended normally after 35 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        14
## 
##                                                   Used       Total
##   Number of observations                          6312        9688
## 
## Model Test User Model:
##                                                       
##   Test statistic                              1325.712
##   Degrees of freedom                                14
##   P-value (Chi-square)                           0.000
## 
## Model Test Baseline Model:
## 
##   Test statistic                             11105.204
##   Degrees of freedom                                21
##   P-value                                        0.000
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.882
##   Tucker-Lewis Index (TLI)                       0.822
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)             -83738.639
##   Loglikelihood unrestricted model (H1)     -83075.783
##                                                       
##   Akaike (AIC)                              167505.277
##   Bayesian (BIC)                            167599.780
##   Sample-size adjusted Bayesian (SABIC)     167555.292
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.122
##   90 Percent confidence interval - lower         0.116
##   90 Percent confidence interval - upper         0.127
##   P-value H_0: RMSEA <= 0.050                    0.000
##   P-value H_0: RMSEA >= 0.080                    1.000
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.060
## 
## Parameter Estimates:
## 
##   Standard errors                             Standard
##   Information                                 Expected
##   Information saturated (h1) model          Structured
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 =~                                                                 
##     in501rer          1.000                               0.980    0.527
##     in502rer          1.362    0.047   28.878    0.000    1.334    0.639
##     in503rer          0.841    0.031   27.584    0.000    0.825    0.562
##   F2 =~                                                                 
##     in504rer   (a)    1.000                               1.392    0.789
##     in505rer   (a)    1.000                               1.392    0.782
##     in506rer          0.705    0.017   42.327    0.000    0.982    0.544
##     in507rer          0.710    0.018   40.436    0.000    0.988    0.523
## 
## Covariances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 ~~                                                                 
##     F2                0.959    0.036   26.914    0.000    0.703    0.703
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .in501rer          2.491    0.054   45.918    0.000    2.491    0.722
##    .in502rer          2.581    0.069   37.432    0.000    2.581    0.592
##    .in503rer          1.474    0.034   43.806    0.000    1.474    0.684
##    .in504rer          1.175    0.031   37.714    0.000    1.175    0.378
##    .in505rer          1.233    0.032   38.626    0.000    1.233    0.389
##    .in506rer          2.287    0.045   50.713    0.000    2.287    0.704
##    .in507rer          2.589    0.050   51.303    0.000    2.589    0.726
##     F1                0.960    0.053   18.193    0.000    1.000    1.000
##     F2                1.937    0.046   41.957    0.000    1.000    1.000
### the chi-square is a little higher than the original two-factor model, the AIC and BIC are a little higher than the original two-factor model, the CFI is a little lower than the original two-factor model but the BIC is a tad higher than the original two-factor model, and the RMSEA is a tad lower. overall, wouldn't say this is a better fit than the original two-factor model, despite gaining one degree of freedom.

# visualization of the model.
semPaths(model_2fEQ_fit)

# take a look at the expected cov matrix and the residual cov matrix.
exp_matrix_2fs <- fitted(model_2fEQ_fit)
exp_matrix_2fs
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  3.451                                          
## in502rer  1.308  4.361                                   
## in503rer  0.808  1.100  2.154                            
## in504rer  0.959  1.306  0.807  3.112                     
## in505rer  0.959  1.306  0.807  1.937  3.170              
## in506rer  0.676  0.921  0.569  1.366  1.366  3.251       
## in507rer  0.681  0.927  0.573  1.375  1.375  0.970  3.565
resid_matrix_2fs <- resid(model_2fEQ_fit)
resid_matrix_2fs
## $type
## [1] "raw"
## 
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  0.000                                          
## in502rer -0.224  0.000                                   
## in503rer  0.059  0.113  0.000                            
## in504rer  0.063  0.160 -0.037 -0.134                     
## in505rer  0.138 -0.102 -0.209  0.103  0.141              
## in506rer  0.085 -0.087 -0.161 -0.305  0.103  0.000       
## in507rer  0.236  0.184 -0.006 -0.322 -0.010  0.746  0.000

two-factor model – vol dropped.

# instead of equating volunteer identity and group/organization identity, i'm going to try dropping volunteer identity from the model to see if that improves fit.
model_2fD <- "
  F1 = ~ in501rer + in502rer + in503rer
  F2 = ~ in505rer + in506rer + in507rer"

# run the CFA and take a look at the fit indices.
model_2fD_fit <- cfa(model_2fD, data = data)
summary(model_2fD_fit, standard = TRUE, fit.measures = TRUE)
## lavaan 0.6.14 ended normally after 39 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                        13
## 
##                                                   Used       Total
##   Number of observations                          6335        9688
## 
## Model Test User Model:
##                                                       
##   Test statistic                               388.494
##   Degrees of freedom                                 8
##   P-value (Chi-square)                           0.000
## 
## Model Test Baseline Model:
## 
##   Test statistic                              7007.870
##   Degrees of freedom                                15
##   P-value                                        0.000
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.946
##   Tucker-Lewis Index (TLI)                       0.898
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)             -73202.829
##   Loglikelihood unrestricted model (H1)     -73008.582
##                                                       
##   Akaike (AIC)                              146431.659
##   Bayesian (BIC)                            146519.459
##   Sample-size adjusted Bayesian (SABIC)     146478.148
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.087
##   90 Percent confidence interval - lower         0.079
##   90 Percent confidence interval - upper         0.094
##   P-value H_0: RMSEA <= 0.050                    0.000
##   P-value H_0: RMSEA >= 0.080                    0.935
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.043
## 
## Parameter Estimates:
## 
##   Standard errors                             Standard
##   Information                                 Expected
##   Information saturated (h1) model          Structured
## 
## Latent Variables:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 =~                                                                 
##     in501rer          1.000                               0.984    0.530
##     in502rer          1.327    0.049   26.928    0.000    1.306    0.625
##     in503rer          0.859    0.032   26.466    0.000    0.845    0.576
##   F2 =~                                                                 
##     in505rer          1.000                               1.168    0.642
##     in506rer          1.073    0.029   37.039    0.000    1.253    0.695
##     in507rer          1.104    0.030   36.901    0.000    1.289    0.683
## 
## Covariances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##   F1 ~~                                                                 
##     F2                0.713    0.032   22.377    0.000    0.620    0.620
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##    .in501rer          2.483    0.056   44.243    0.000    2.483    0.719
##    .in502rer          2.656    0.073   36.274    0.000    2.656    0.609
##    .in503rer          1.440    0.035   40.798    0.000    1.440    0.668
##    .in505rer          1.949    0.047   41.374    0.000    1.949    0.588
##    .in506rer          1.683    0.046   36.255    0.000    1.683    0.517
##    .in507rer          1.905    0.051   37.511    0.000    1.905    0.534
##     F1                0.968    0.055   17.629    0.000    1.000    1.000
##     F2                1.364    0.057   23.794    0.000    1.000    1.000
### chi-square is still significant, but again, we have a large sample size. the AIC and BIC are quite a bit lower than the original two-factor model. the CFI does not reach the 0.95 threshold but is close! the TLI still does not reach the 0.95 threshold. the RMSEA still does not fall below the 0.06 threshold. this model fits a bit better than the others, and is probably the best fit overall.

# visualization of the model.
semPaths(model_2fD_fit)

# take a look at the expected cov matrix and the residual cov matrix.
exp_matrix_2fd <- fitted(model_2fD_fit)
exp_matrix_2fd
## $cov
##          in501r in502r in503r in505r in506r in507r
## in501rer  3.451                                   
## in502rer  1.285  4.362                            
## in503rer  0.832  1.104  2.154                     
## in505rer  0.713  0.946  0.612  3.313              
## in506rer  0.765  1.015  0.657  1.463  3.253       
## in507rer  0.787  1.045  0.676  1.506  1.615  3.568
resid_matrix_2fd <- resid(model_2fD_fit)
resid_matrix_2fd
## $type
## [1] "raw"
## 
## $cov
##          in501r in502r in503r in505r in506r in507r
## in501rer  0.000                                   
## in502rer -0.199  0.000                            
## in503rer  0.035  0.106  0.000                     
## in505rer  0.382  0.259 -0.017  0.000              
## in506rer -0.001 -0.177 -0.249  0.009  0.000       
## in507rer  0.131  0.066 -0.111 -0.137  0.102  0.000

null model.

# lastly, just gonna take a look at the null model to cover all bases.
null_model <- "
  in501rer ~~ in501rer
  in502rer ~~ in502rer
  in503rer ~~ in503rer
  in504rer ~~ in504rer
  in505rer ~~ in505rer
  in506rer ~~ in506rer
  in507rer ~~ in507rer"

# run the CFA and take a look at the fit indices.
null_model_fit <- cfa(null_model, data = data)
summary(null_model_fit, standard = TRUE, fit.measures = TRUE)
## lavaan 0.6.14 ended normally after 9 iterations
## 
##   Estimator                                         ML
##   Optimization method                           NLMINB
##   Number of model parameters                         7
## 
##                                                   Used       Total
##   Number of observations                          6312        9688
## 
## Model Test User Model:
##                                                        
##   Test statistic                              11105.204
##   Degrees of freedom                                 21
##   P-value (Chi-square)                            0.000
## 
## Model Test Baseline Model:
## 
##   Test statistic                             11105.204
##   Degrees of freedom                                21
##   P-value                                        0.000
## 
## User Model versus Baseline Model:
## 
##   Comparative Fit Index (CFI)                    0.000
##   Tucker-Lewis Index (TLI)                       0.000
## 
## Loglikelihood and Information Criteria:
## 
##   Loglikelihood user model (H0)             -88628.385
##   Loglikelihood unrestricted model (H1)     -83075.783
##                                                       
##   Akaike (AIC)                              177270.770
##   Bayesian (BIC)                            177318.022
##   Sample-size adjusted Bayesian (SABIC)     177295.778
## 
## Root Mean Square Error of Approximation:
## 
##   RMSEA                                          0.289
##   90 Percent confidence interval - lower         0.285
##   90 Percent confidence interval - upper         0.294
##   P-value H_0: RMSEA <= 0.050                    0.000
##   P-value H_0: RMSEA >= 0.080                    1.000
## 
## Standardized Root Mean Square Residual:
## 
##   SRMR                                           0.300
## 
## Parameter Estimates:
## 
##   Standard errors                             Standard
##   Information                                 Expected
##   Information saturated (h1) model          Structured
## 
## Variances:
##                    Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
##     in501rer          3.451    0.061   56.178    0.000    3.451    1.000
##     in502rer          4.361    0.078   56.178    0.000    4.361    1.000
##     in503rer          2.154    0.038   56.178    0.000    2.154    1.000
##     in504rer          2.978    0.053   56.178    0.000    2.978    1.000
##     in505rer          3.310    0.059   56.178    0.000    3.310    1.000
##     in506rer          3.251    0.058   56.178    0.000    3.251    1.000
##     in507rer          3.565    0.063   56.178    0.000    3.565    1.000
### chi-square is much higher than the original two-factor model and we lost 6 degrees of freedom. AIC and BIC are higher than the original two-factor model, CFI and TLI are 0, and RMSEA is much higher than the original two-factor model.

# visualization of the model.
semPaths(null_model_fit)

# take a look at the expected cov matrix and the residual cov matrix.
exp_matrix_null <- fitted(null_model_fit)
exp_matrix_null
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  3.451                                          
## in502rer  0.000  4.361                                   
## in503rer  0.000  0.000  2.154                            
## in504rer  0.000  0.000  0.000  2.978                     
## in505rer  0.000  0.000  0.000  0.000  3.310              
## in506rer  0.000  0.000  0.000  0.000  0.000  3.251       
## in507rer  0.000  0.000  0.000  0.000  0.000  0.000  3.565
resid_matrix_null <- resid(null_model_fit)
resid_matrix_null
## $type
## [1] "raw"
## 
## $cov
##          in501r in502r in503r in504r in505r in506r in507r
## in501rer  0.000                                          
## in502rer  1.083  0.000                                   
## in503rer  0.867  1.213  0.000                            
## in504rer  1.022  1.466  0.769  0.000                     
## in505rer  1.097  1.204  0.598  2.039  0.000              
## in506rer  0.762  0.834  0.409  1.061  1.470  0.000       
## in507rer  0.917  1.111  0.567  1.053  1.365  1.716  0.000