lavaan WARNING: some models are based on a different set of observed variables

[Leslie Rutkowski]

Hi all,

I'm putting together an unconditional vs. conditional growth curve model example and the two models - that only differ in terms of the time invariant covariates - produce the "different set of observed variables" error. But there is no missing data on the covariates. And I am dealing with nested models (unconditional is nested within conditional). Also, the sample sizes are the same between the two models. I can't figure out what might be generating this error. When I comment out the covariates, the model fit is identical.

Here is a replicable example: 

library(lavaan)

library(reshape2)

#--------------------------

# Using Singer & Willett tolerance 

# toward deviant behavior data

#--------------------------

# Wide data

tolerance <- read.table("https://stats.idre.ucla.edu/wp-content/uploads/2016/02/tolerance1.txt", sep=",", header=T)

#------------------------

# U-LGCM

ulgcm <- 'i =~ 1*tol11 + 1*tol12 + 1*tol13 + 1*tol14 + 1*tol15

          s =~ 0*tol11 + 1*tol12 + 2*tol13 + 3*tol14 + 4*tol15 

          tol11 ~~ a*tol11

          tol12 ~~ a*tol12

          tol13 ~~ a*tol13

          tol14 ~~ a*tol14

          tol15 ~~ a*tol15

        '

# Model fit and summary

ulgcm.fit <- growth(ulgcm, data = tolerance)

summary(ulgcm.fit)

#------------------------

#------------------------

# C-LGCM

clgcm <- '# The "growth" model

          i =~ 1*tol11 + 1*tol12 + 1*tol13 + 1*tol14 + 1*tol15

          s =~ 0*tol11 + 1*tol12 + 2*tol13 + 3*tol14 + 4*tol15 

          tol11 ~~ a*tol11

          tol12 ~~ a*tol12

          tol13 ~~ a*tol13

          tol14 ~~ a*tol14

          tol15 ~~ a*tol15

          # Model to predict growth

          i ~ male

          s ~ male

          i ~~ s

'

# Model fit and summary

clgcm.fit <- growth(clgcm, data = tolerance)

summary(clgcm.fit)

# Model comparison

lavTestLRT(ulgcm.fit, clgcm.fit)

#------------------------

[Yves Rosseel]

to make the models nested, you should add the covariates in the
unconditional model too, but fix the regression slopes to zero:

i ~ 0*male
s ~ 0*male

Yves.

> --
> You received this message because you are subscribed to the Google
> Groups "lavaan" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to lavaan+un...@googlegroups.com
> <mailto:lavaan+un...@googlegroups.com>.
> To post to this group, send email to lav...@googlegroups.com
> <mailto:lav...@googlegroups.com>.
> Visit this group at https://groups.google.com/group/lavaan.
> For more options, visit https://groups.google.com/d/optout.