The Cultural Fairness of the 12-Item General Health Questionnaire Among Diverse Adolescents

Supplemental Materials

The Cultural Fairness of the 12-Item General Health Questionnaire Among Diverse Adolescents

by A. Bowe, 2016, Psychological Assessment

MPLUS 7.0 Testing for Configural Invariance

TITLE:

test for configural invariance

DATA:

FILE IS C:\Name.dat;

FORMAT IS free;

VARIABLE:

NAMES ARE id clus w2 sex ethnic c1-c12;

MISSING ARE all (-99);

CATEGORICAL ARE c1 - c12;

idvariable are id;

cluster = clus;

weight = w2;

usevariables are c1 - c12;

useobservations ARE (ethnic < 8);!if i want to pick ethnic groups

Grouping IS ethnic (1 = W 2 = M 3 = I 4 = P 5 = B 6 = C 7 = A );

ANALYSIS:

Estimator = ML; type = complex mgroup ; !By default it did WLSMV

MODEL is configural; !for confirmatory analysis in CFA framework

!Baseline model

MODEL:

PoorHealth BY c1- c12;

c2 c5 c6 c9 c10 c11 WITH c2 c5 c6 c9 c10 c11;

OUTPUT: standardized MODINDICES (ALL 3.84);

Savedata:

File="Name.xls";

R Syntax: DIF ANALYSIS

#Read data in#

#evoke lordif package

library(lordif)

data.mental<-read.table("FILE.dat", header = TRUE, na.strings = c("-99"))

#changes ethnicity to factor format from vector of integers

data.mental$Race.c<-factor(data.mental$Race, labels=c("White","Mixed","Indian","Pakistani",

"Bangladeshi", "Caribbean", "African", "Other"))

##### drop the OTHER category

data.mental<-subset(data.mental, Race <8)

dim(data.mental)

data.mental$Race.c<-factor(data.mental$Race.c)

str(data.mental$Race.c)

#Create matrix for response variable

Resp<-data.mental[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-data.mental[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

is.factor(Ethnicity) ## this checks to make sure is really is a factor

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = 0.05/77, minCell = 5)

# bonferonni correction of .05/11*7

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("White", "Mixed","Indian","Pakis","Bang", "Caribbean", "African"))

## Error messages. Consulted with author of package and completed the syntax below###

______

### Do with 2 groups at a time, because 7 couldn't run###

#### Carib and African #################

Carib<-subset(data.mental, Race == 6)

African<-subset(data.mental, Race == 7)

Black<-rbind(Carib,African)

dim(Black)

summary(Black)

Black$Race.c<-factor(Black$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-Black[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-Black[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

is.factor(Ethnicity) ## this checks to make sure is really is a factor

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0045, minCell = 5)

# bonferonni correction of .05/11

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("Caribbean", "African"))

##Two items flagged for DIF, (2,11)

############## ASIAN GROUPS ###############

Pak<-subset(data.mental, Race == 4)

Ind<-subset(data.mental, Race == 3)

Bang<-subset(data.mental, Race == 5)

Asian<-rbind(Pak, Ind, Bang)

dim(Asian)

summary(Asian)

Asian$Race.c<-factor(Asian$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-Asian[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-Asian[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

is.factor(Ethnicity) ## this checks to make sure is really is a factor

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0022, minCell = 5)

# bonferonni correction of .05/22

print(ethnicDIF)

summary(ethnicDIF)

## Results no item flagged for DIF between these Asian groups

###################### Mixed group ##############

Mix<-subset(data.mental, Race == 2)

White<-subset(data.mental, Race == 1)

Mix_White<-rbind(Mix,White)

dim(Mix_White)

Mix_White$Race.c<-factor(Mix_White$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-Mix_White[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-Mix_White[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

is.factor(Ethnicity) ## this checks to make sure is really is a factor

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0045, minCell = 5)

# bonferonni correction of .05/11

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("White", "Mixed"))

## no items flagged for DIF between white & mixed

################# WHITE TO 3 ASIAN GROUPS#################################

Asian$together<- 9

White$together <-1

Asian$together<-as.factor(Asian$together)

White$together<-as.factor(White$together)

Asian_White<-rbind(Asian,White)

dim(Asian_White)

Asian_White$Race.c<-factor(Asian_White$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-Asian_White[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-Asian_White[19] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

Ethnicity<-as.factor(Ethnicity) ## to drop the other ethnicities with 0 count

Ethnicity<-factor(Ethnicity, labels=c("Asian","White"))

is.factor(Ethnicity) ## this checks to make sure is really is a factor

str(Ethnicity)

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0045, minCell = 5)

# bonferonni correction of .05/11

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("Asian", "White"))

## Iteration 3 : 7 items flagged for DIF (1,2,4,5,7,8,9)

######################## white to carib group #################################

Carib$together<-6

Carib_White<-rbind(Carib,White)

dim(Carib_White)

Carib_White$Race.c<-factor(Carib_White$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-Carib_White[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-Carib_White[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

Ethnicity<-as.factor(Ethnicity) ## to drop the other ethnicities with 0 count

Ethnicity<-factor(Ethnicity, labels=c("White","Carib"))

is.factor(Ethnicity) ## this checks to make sure is really is a factor

str(Ethnicity)

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0045, minCell = 5)

# bonferonni correction of .05/11

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("White", "Carib"))

Item 1 flagged for DIF (Item 1)

######################## white to african group ##############################

African$together<-7

African_White<-rbind(African,White)

dim(African_White)

African_White$Race.c<-factor(African_White$Race.c) # this drops the factor levels that are no longer there.

#Needed to run lordif

#Create matrix for response variable

Resp<-African_White[c(6:17)]

Resp<-as.matrix(Resp)

is.matrix(Resp)

#Create FACTOR vector for categorical variable

Ethnicity<-African_White[18] ##this grabs the column of groups. The grabbing puts it back into list form

Ethnicity<-unlist(Ethnicity) ## unlist changes it to factor format

Ethnicity<-as.factor(Ethnicity) ## to drop the other ethnicities with 0 count

Ethnicity<-factor(Ethnicity, labels=c("White","African"))

is.factor(Ethnicity) ## this checks to make sure is really is a factor

str(Ethnicity)

dim(Ethnicity);dim(Resp)

summary(Resp);summary(Ethnicity)

ethnicDIF<-lordif(Resp, Ethnicity, criterion = "Chisqr", alpha = .0045, minCell = 5)

# bonferonni correction of .05/11

print(ethnicDIF)

summary(ethnicDIF)

par(mfrow=c(3,2),pty="s")

plot(ethnicDIF, labels = c("White", "African"))

Iteration 4 : 7 items flagged for DIF (1,3,4,7,8,11,12)

###### SCALE LEVEL BIAS ###############

setwd("Filepath")

library(xlsx)

# NEED TO OBTAIN ITEM PARAMETERS #

library(ltm)

# First, create a dataframe with just manifest variables

white_DTF<-subset(data.mental, Race == 1, select = W2concenYP:W2happyYP)

mixed_DTF<-subset(data.mental, Race == 2, select = W2concenYP: W2happyYP)

indian_DTF<-subset(data.mental, Race == 3, select = W2concenYP: W2happyYP)

pak_DTF<-subset(data.mental, Race == 4, select = W2concenYP: W2happyYP)

bang_DTF<-subset(data.mental, Race == 5, select = W2concenYP: W2happyYP)

carib_DTF<-subset(data.mental, Race == 6, select = W2concenYP: W2happyYP)

african_DTF<-subset(data.mental, Race == 7, select = W2concenYP: W2happyYP)

asian_DTF<-rbind(indian_DTF,pak_DTF,bang_DTF)

dim(asian_DTF)

# Second, obtain item parameters #

#grm(data, constrained = FALSE, IRT.param = TRUE, Hessian = FALSE,

#start.val = NULL, na.action = NULL, control = list())

itempar_white<-grm(white_DTF[c(1:12)], constrained = FALSE, IRT.param = TRUE, Hessian = FALSE, start.val = NULL, na.action = NULL, control = list())

itempar_mix<-grm(mixed_DTF[c(1:12)], constrained = FALSE, IRT.param = TRUE, Hessian = FALSE, start.val = NULL, na.action = NULL, control = list())

itempar_carib<-grm(carib_DTF[c(1:12)], constrained = FALSE, IRT.param = TRUE, Hessian = FALSE, start.val = NULL, na.action = NULL, control = list())

itempar_afr<-grm(african_DTF[c(1:12)], constrained = FALSE, IRT.param = TRUE, Hessian = FALSE, start.val = NULL, na.action = NULL, control = list())

itempar_asian<-grm(asian_DTF[c(1:12)], constrained = FALSE, IRT.param = TRUE, Hessian = FALSE, start.val = NULL, na.action = NULL, control = list())

#EXTRACT ITEM PARAMETERS

ipar_white<-extract(itempar_white)

ipar_mixed<-extract(itempar_mix)

ipar_carib<-extract(itempar_carib)

ipar_afr<-extract(itempar_afr)

ipar_asian<-extract(itempar_asian)

## OBTAIN THETA ESTIMATES FOR ALL GROUPS ###

theta <- seq(-4,4,.1)

##calctheta(ipar, resp.data, theta, prior.mean = 0, prior.sd = 1) #Calculates the Expected A Posteriori (EAP) theta and the associated posterior standard deviation(PSD) as standard error estimates.

#WHITE

theta_white<-calctheta(ipar_white, white_DTF, theta,prior.mean = 0, prior.sd = 1)

write.xlsx(theta_white, "thetascores_white.xlsx")

write.xlsx(white_DTF, "rawscores_white.xlsx")

#MIXED

theta_mixed<-calctheta(ipar_mixed, mixed_DTF, theta,prior.mean = 0, prior.sd = 1)

str(theta_mixed)

write.xlsx(mixed_DTF, "rawscores_mixed.xlsx")

write.xlsx(theta_mixed, "thetascores_mixed.xlsx")

#CARIBBEAN

theta_carib<-calctheta(ipar_carib, carib_DTF, theta,prior.mean = 0, prior.sd = 1)

str(theta_carib)

write.xlsx(carib_DTF, "rawscores_carib.xlsx")

write.xlsx(theta_carib, "thetascores_carib.xlsx")

#AFRICAN

theta_afr<-calctheta(ipar_afr, african_DTF, theta,prior.mean = 0, prior.sd = 1)

str(theta_afr)

write.xlsx(african_DTF, "rawscores_afr.xlsx")

write.xlsx(theta_afr, "thetascores_afr.xlsx")

#ASIAN

theta_asian<-calctheta(ipar_asian, asian_DTF, theta,prior.mean = 0, prior.sd = 1)

str(theta_asian)

write.xlsx(asian_DTF, "rawscores_asian.xlsx")

write.xlsx(theta_asian, "thetascores_asian.xlsx")

##### NOW IMPORT OBJECTS AND DATAFRAMES INTO VISUAL AID #####