Dissertation Analyses

load(file="/Users/meganwilliams/Desktop/Dissertation/StroopMixed.rdata")
load(file="/Users/meganwilliams/Desktop/Dissertation/Allvars.rdata")
library(effects)
library(interactions)
library(rcompanion)
library(car)

Trail Making Test Part B - Psychological Aggression

Model 1

options(scipen = 999)
TMTBlog1 <- glm(PsychAggress ~ TrailsB + WRATtotal, data=Allvars,family = "binomial")
summary(TMTBlog1)
## 
## Call:
## glm(formula = PsychAggress ~ TrailsB + WRATtotal, family = "binomial", 
##     data = Allvars)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -2.1102   0.4895   0.5148   0.5478   0.7594  
## 
## Coefficients:
##              Estimate Std. Error z value Pr(>|z|)
## (Intercept)  1.339399   0.816391   1.641    0.101
## TrailsB     -0.001334   0.001663  -0.802    0.422
## WRATtotal    0.015670   0.016401   0.955    0.339
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 505.4  on 640  degrees of freedom
## Residual deviance: 502.6  on 638  degrees of freedom
## AIC: 508.6
## 
## Number of Fisher Scoring iterations: 4
exp(cbind(OR = coef(TMTBlog1), confint(TMTBlog1)))
##                    OR     2.5 %    97.5 %
## (Intercept) 3.8167494 0.7865987 19.464758
## TrailsB     0.9986672 0.9955020  1.002036
## WRATtotal   1.0157937 0.9832268  1.048699
########Compare to null model 
#Difference in Deviance
with(TMTBlog1,null.deviance - deviance)
## [1] 2.79723
#Degrees of freedom for the difference between two models
with(TMTBlog1,df.null - df.residual)
## [1] 2
#p-value
with(TMTBlog1,pchisq(null.deviance-deviance,df.null-df.residual,lower.tail=FALSE))
## [1] 0.2469387
#Pseudo R-Squared
nagelkerke(TMTBlog1)
## $Models
##                                                                    
## Model: "glm, PsychAggress ~ TrailsB + WRATtotal, binomial, Allvars"
## Null:  "glm, PsychAggress ~ 1, binomial, Allvars"                  
## 
## $Pseudo.R.squared.for.model.vs.null
##                              Pseudo.R.squared
## McFadden                           0.00553467
## Cox and Snell (ML)                 0.00435435
## Nagelkerke (Cragg and Uhler)       0.00798297
## 
## $Likelihood.ratio.test
##  Df.diff LogLik.diff  Chisq p.value
##       -2     -1.3986 2.7972 0.24694
## 
## $Number.of.observations
##           
## Model: 641
## Null:  641
## 
## $Messages
## [1] "Note: For models fit with REML, these statistics are based on refitting with ML"
## 
## $Warnings
## [1] "None"

Model 3

options(scipen = 999)
TMTBlog3 <- glm(PsychAggress ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal, data = Allvars, family = "binomial")
summary(TMTBlog3)
## 
## Call:
## glm(formula = PsychAggress ~ (TrailsB + Sex + PovStat)^3 + Age + 
##     WRATtotal, family = "binomial", data = Allvars)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -2.3166   0.4161   0.4923   0.5692   0.9530  
## 
## Coefficients:
##                                Estimate  Std. Error z value Pr(>|z|)  
## (Intercept)                  2.11869216  1.06981158   1.980   0.0477 *
## TrailsB                      0.00006311  0.00289151   0.022   0.9826  
## SexMen                      -0.28635201  0.49006156  -0.584   0.5590  
## PovStatBelow                 0.27642697  0.71307763   0.388   0.6983  
## Age                         -0.02183121  0.01355727  -1.610   0.1073  
## WRATtotal                    0.02252136  0.01718596   1.310   0.1900  
## TrailsB:SexMen              -0.00202431  0.00347289  -0.583   0.5600  
## TrailsB:PovStatBelow        -0.00157107  0.00516673  -0.304   0.7611  
## SexMen:PovStatBelow         -0.88586893  1.02357774  -0.865   0.3868  
## TrailsB:SexMen:PovStatBelow  0.01099893  0.00806233   1.364   0.1725  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 505.40  on 640  degrees of freedom
## Residual deviance: 490.99  on 631  degrees of freedom
## AIC: 510.99
## 
## Number of Fisher Scoring iterations: 5
exp(cbind(OR = coef(TMTBlog3), confint(TMTBlog3)))
##                                    OR      2.5 %    97.5 %
## (Intercept)                 8.3202488 1.04184791 69.643273
## TrailsB                     1.0000631 0.99469992  1.006199
## SexMen                      0.7509982 0.28438310  1.957070
## PovStatBelow                1.3184107 0.33189532  5.588898
## Age                         0.9784054 0.95254150  1.004651
## WRATtotal                   1.0227769 0.98855518  1.057632
## TrailsB:SexMen              0.9979777 0.99095614  1.004673
## TrailsB:PovStatBelow        0.9984302 0.98861094  1.009336
## SexMen:PovStatBelow         0.4123557 0.05289882  3.010741
## TrailsB:SexMen:PovStatBelow 1.0110596 0.99573384  1.028653
########Compare to null model 
#Difference in Deviance
with(TMTBlog3,null.deviance - deviance)
## [1] 14.41466
#Degrees of freedom for the difference between two models
with(TMTBlog3,df.null - df.residual)
## [1] 9
#p-value
with(TMTBlog3,pchisq(null.deviance-deviance,df.null-df.residual,lower.tail=FALSE))
## [1] 0.1083207
#Pseudo R-Squared
nagelkerke(TMTBlog3)
## $Models
##                                                                                              
## Model: "glm, PsychAggress ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal, binomial, Allvars"
## Null:  "glm, PsychAggress ~ 1, binomial, Allvars"                                            
## 
## $Pseudo.R.squared.for.model.vs.null
##                              Pseudo.R.squared
## McFadden                            0.0285212
## Cox and Snell (ML)                  0.0222368
## Nagelkerke (Cragg and Uhler)        0.0407675
## 
## $Likelihood.ratio.test
##  Df.diff LogLik.diff  Chisq p.value
##       -9     -7.2073 14.415 0.10832
## 
## $Number.of.observations
##           
## Model: 641
## Null:  641
## 
## $Messages
## [1] "Note: For models fit with REML, these statistics are based on refitting with ML"
## 
## $Warnings
## [1] "None"

Compare Models 1 & 3

anova(TMTBlog1,TMTBlog3,test = "LR")
## Analysis of Deviance Table
## 
## Model 1: PsychAggress ~ TrailsB + WRATtotal
## Model 2: PsychAggress ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal
##   Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1       638     502.60                     
## 2       631     490.99  7   11.617   0.1139

Trail Making Test Part B - Physical Assault

Model 1

options(scipen = 999)
TMTBlog1 <- glm(PhysAssault ~ TrailsB + WRATtotal, data=Allvars,family = "binomial")
summary(TMTBlog1)
## 
## Call:
## glm(formula = PhysAssault ~ TrailsB + WRATtotal, family = "binomial", 
##     data = Allvars)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.5637  -0.5337  -0.5159  -0.4854   2.2079  
## 
## Coefficients:
##               Estimate Std. Error z value Pr(>|z|)  
## (Intercept) -2.2512544  0.9100576  -2.474   0.0134 *
## TrailsB     -0.0009631  0.0019131  -0.503   0.6147  
## WRATtotal    0.0092431  0.0181183   0.510   0.6099  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 482.53  on 640  degrees of freedom
## Residual deviance: 481.59  on 638  degrees of freedom
## AIC: 487.59
## 
## Number of Fisher Scoring iterations: 4
exp(cbind(OR = coef(TMTBlog1), confint(TMTBlog1)))
##                    OR      2.5 %    97.5 %
## (Intercept) 0.1052671 0.01689774 0.6044881
## TrailsB     0.9990374 0.99509995 1.0026319
## WRATtotal   1.0092860 0.97468590 1.0466183
########Compare to null model 
#Difference in Deviance
with(TMTBlog1,null.deviance - deviance)
## [1] 0.941138
#Degrees of freedom for the difference between two models
with(TMTBlog1,df.null - df.residual)
## [1] 2
#p-value
with(TMTBlog1,pchisq(null.deviance-deviance,df.null-df.residual,lower.tail=FALSE))
## [1] 0.6246467
#Pseudo R-Squared
nagelkerke(TMTBlog1)
## $Models
##                                                                   
## Model: "glm, PhysAssault ~ TrailsB + WRATtotal, binomial, Allvars"
## Null:  "glm, PhysAssault ~ 1, binomial, Allvars"                  
## 
## $Pseudo.R.squared.for.model.vs.null
##                              Pseudo.R.squared
## McFadden                           0.00195041
## Cox and Snell (ML)                 0.00146716
## Nagelkerke (Cragg and Uhler)       0.00277374
## 
## $Likelihood.ratio.test
##  Df.diff LogLik.diff   Chisq p.value
##       -2    -0.47057 0.94114 0.62465
## 
## $Number.of.observations
##           
## Model: 641
## Null:  641
## 
## $Messages
## [1] "Note: For models fit with REML, these statistics are based on refitting with ML"
## 
## $Warnings
## [1] "None"

Model 3

options(scipen = 999)
TMTBlog3 <- glm(PhysAssault ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal, data = Allvars, family = "binomial")
summary(TMTBlog3)
## 
## Call:
## glm(formula = PhysAssault ~ (TrailsB + Sex + PovStat)^3 + Age + 
##     WRATtotal, family = "binomial", data = Allvars)
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.8076  -0.5530  -0.4680  -0.3874   2.3898  
## 
## Coefficients:
##                               Estimate Std. Error z value Pr(>|z|)  
## (Intercept)                 -1.3070879  1.1474161  -1.139    0.255  
## TrailsB                     -0.0003952  0.0033321  -0.119    0.906  
## SexMen                      -0.4164733  0.5404079  -0.771    0.441  
## PovStatBelow                 0.3301475  0.5989564   0.551    0.581  
## Age                         -0.0339108  0.0142097  -2.386    0.017 *
## WRATtotal                    0.0193090  0.0188812   1.023    0.306  
## TrailsB:SexMen               0.0022476  0.0043351   0.518    0.604  
## TrailsB:PovStatBelow         0.0012567  0.0049259   0.255    0.799  
## SexMen:PovStatBelow          0.5698502  0.9182035   0.621    0.535  
## TrailsB:SexMen:PovStatBelow -0.0046844  0.0073815  -0.635    0.526  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 482.53  on 640  degrees of freedom
## Residual deviance: 468.86  on 631  degrees of freedom
## AIC: 488.86
## 
## Number of Fisher Scoring iterations: 5
exp(cbind(OR = coef(TMTBlog3), confint(TMTBlog3)))
##                                    OR      2.5 %     97.5 %
## (Intercept)                 0.2706069 0.02769454  2.5163377
## TrailsB                     0.9996049 0.99235430  1.0056459
## SexMen                      0.6593681 0.22560150  1.8962614
## PovStatBelow                1.3911733 0.42577175  4.5105093
## Age                         0.9666577 0.93969358  0.9936472
## WRATtotal                   1.0194966 0.98316810  1.0589190
## TrailsB:SexMen              1.0022501 0.99370070  1.0110710
## TrailsB:PovStatBelow        1.0012575 0.99134463  1.0110721
## SexMen:PovStatBelow         1.7680022 0.29233481 10.8531448
## TrailsB:SexMen:PovStatBelow 0.9953265 0.98043330  1.0096599
########Compare to null model 
#Difference in Deviance
with(TMTBlog3,null.deviance - deviance)
## [1] 13.67086
#Degrees of freedom for the difference between two models
with(TMTBlog3,df.null - df.residual)
## [1] 9
#p-value
with(TMTBlog3,pchisq(null.deviance-deviance,df.null-df.residual,lower.tail=FALSE))
## [1] 0.1345244
#Pseudo R-Squared
nagelkerke(TMTBlog3)
## $Models
##                                                                                             
## Model: "glm, PhysAssault ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal, binomial, Allvars"
## Null:  "glm, PhysAssault ~ 1, binomial, Allvars"                                            
## 
## $Pseudo.R.squared.for.model.vs.null
##                              Pseudo.R.squared
## McFadden                            0.0283315
## Cox and Snell (ML)                  0.0211016
## Nagelkerke (Cragg and Uhler)        0.0398937
## 
## $Likelihood.ratio.test
##  Df.diff LogLik.diff  Chisq p.value
##       -9     -6.8354 13.671 0.13452
## 
## $Number.of.observations
##           
## Model: 641
## Null:  641
## 
## $Messages
## [1] "Note: For models fit with REML, these statistics are based on refitting with ML"
## 
## $Warnings
## [1] "None"

Compare Models 1 & 3

anova(TMTBlog1,TMTBlog3,test = "LR")
## Analysis of Deviance Table
## 
## Model 1: PhysAssault ~ TrailsB + WRATtotal
## Model 2: PhysAssault ~ (TrailsB + Sex + PovStat)^3 + Age + WRATtotal
##   Resid. Df Resid. Dev Df Deviance Pr(>Chi)  
## 1       638     481.59                       
## 2       631     468.86  7    12.73  0.07897 .
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1