Re: Question on the bifactor invariance modeling

[Phil Chalmers]

Hi Minji,

This looks correct. Standardized loadings will only be identical when the latent variances are equal, which they are not here. The coef()s will be the same though, which is what is being constrained. HTH.

Phil

On Fri, Jun 24, 2016 at 2:08 PM, Lee, Minji K., M.S., Ph.D. <Lee....@mayo.edu> wrote:

 

Hi Phil,

 

I noticed that when I tested strong factorial invariance (equal slopes and intercepts) using bfactor function, the results indicated only the group factor loadings were set equal between 2 groups. General factor loadings were different. Is this normal? Please see the screenshots below. I didn’t print out all the items. I thought that the highlighted part should have taken care of the issue, but apparently it did not. The code I used is as follows.

 

 

 

group_factor <- c(1,1,1,1,1,1,1,1,1,1,

                  1,1,1,1,1,1,1,2,2,2,

                  2,2,2,2,2,2,2,2,2,2,

                  2,2,2,2,2,2,2,2,2,2,

                  2,2,2,2,2,2,1,1,2,2,

                  1,2,2,2,2,2,2,1,1,1,

                  1,1,1,1,1,1,1,1,1,1,

                  1,1,1,2,2,2,2,2,2,2,

                  2,2,2,2,2,2,2,2,2,2,

                  2,1,1,1,1)

 

 

inv.model2 <- bfactor(combined.data,

                      model=group_factor,

                      invariance= c("slopes", "intercepts","free_var","free_means"), 

                      itemtype="graded",

                      group=grp,

                      quadpts = 32,

                      verbose=TRUE,

                      TOL=1E-05,

                      technical=list(NCYCLES=5000))

 

 

Thank you for your consideration!

Minji

 

Group 1

 

Here is group 2