Measurement invariance - partial invariance

110 views
Skip to first unread message

Isabella Edward

unread,
Jul 30, 2018, 1:01:16 PM7/30/18
to lavaan
Hello guys, looking for some help while beeing complete overstrained with my data. I developed a new questionnaire run a EFA that indicates one latent factor for my 15 Items. Run a CFA that says the one factor solution fits horrible. Allowed for some crossloadings -> fit is for my work acceptable. (The questionnaire doesn't seem to work).

So here's my model + I like to test for measurement invariance across gender (SD01)

library(lavaan)
library(semTools)
#Messinvarianz

Hardiness.1Faktor <-'Hard =~ 1*CH01_01 + CH02_02 + CH04_04 + CH05_05 + CH07_07 + KO01_01 + KO03_03 + KO04_04 + KO05_05 + KO08_08 + CO01_01 + CO03_03 + CO04_04 + CO05_05 + CO09_01
         KO01_01 ~~ KO03_03
         KO01_01 ~~ KO05_05
         CH02_02 ~~ CH04_04
         CH01_01 ~~ KO03_03
         KO01_01 ~~ CO09_01'
fit.1 <- cfa(Hardiness.1Faktor, data=Daten_, estimator ="mlr", std.lv=TRUE, group="SD01")
measurementInvariance(model=fit.1, data=Daten_)

which ends up in the error warning:
Configural model did not converge.

Can anybody tell me how i can fix this problem or whats the source of error?

Terrence Jorgensen

unread,
Jul 31, 2018, 8:14:46 AM7/31/18
to lavaan
Can anybody tell me how i can fix this problem or whats the source of error? 

You told the function that the model syntax (model=) was a fitted lavaan object (fit.1) instead of the model syntax (Hardiness.1Faktor)

fit.1 <- cfa(Hardiness.1Faktor, data=Daten_, estimator ="mlr", std.lv=TRUE, group="SD01")
measurementInvariance
(model=fit.1, data=Daten_)


Instead, pass your syntax, and remember to pass the other lavaan options (you at least need to tell the function what the grouping variable is).

measurementInvariance(model=Hardiness.1Faktor, data=Daten_, estimator ="mlr", std.lv=TRUE, group="SD01")

Terrence D. Jorgensen
Postdoctoral Researcher, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam


Reply all
Reply to author
Forward
0 new messages