0

I'm trying to fit a model with ordinal variables and it seems like, due to a sample size of just N=111, there are quite a lot of empty cells in the correlation tables:

fit <- cfa( model = my.model, data = items, ordered = c("oc1","oc2","oc3","oc4","oc5","oc6","oc7","oc8","oc9","oc10","oc11","oc12","oc13","oc14","oc15","oc16","oc17","oc18","oc19","oc20","oc21","oc22","oc23"), estimator = "WLSMV" )

lavaan WARNING: 253 bivariate tables have empty cells

So I read that lavaan has the zero.add option, but when I pass zero.add = c(0.5, 0.5) to cfa(), I still get the same warning. After inspecting the correlation tables, nothing seems to have changed. The robust fit measures are still not calculated, when inspecting them they are set to NA. Is this normal behavior, or is there something I'm missing? Is there a way to set the option so that cfa() accepts it? I also tried it with lavaan() and the defaults used by cfa() but still nothing…

My version is 0.5.23.1097

Cœur
  • 37,241
  • 25
  • 195
  • 267
Tim Ka
  • 142
  • 2
  • 8

1 Answers1

0

Ok, I found out that WLSMV doesn't even compute the "robust" fit measures and that I'm supposed to use the "scaled" ones. Also, I seem to have overlooked, that setting zero.add to TRUE seems to change fit measures a little bit.

Tim Ka
  • 142
  • 2
  • 8