Extracting baseline (null model) parameters from fitted cfa object?

[Allie Choate]

Hi all,

I posted a few weeks a go on a similar issue of getting NAs for my CFI/TLI in the newest version of Lavaan.

I went back to an older lavaan (version .5-23) and was able to successfully get CFI/TLI values using ML estimation. 

However, when I tried to use MLR, the model did not converge (said matrix was singular, although I have checked this in other ways and it doesn't appear to be...although some variables are on the colinear side, as my model is a longitudinal bifactor). 

It's odd to see a difference between the older and newer Lavaan, but my guess is maybe the independence modeling is not converging properly, and thus my other fit indices are able to be calculated, just not CFI/TLI.

Does anyone have any insight on if there's a way to 'extract' the baseline model that Lavaan is running in the background?

I know you can calculate your own fit indices using your own specified 'baseline model', which I've considered, though when I try to do that it gives me different estimates than were originally computed in the older version of lavaan.

Thanks in advance!

Allie

[Terrence Jorgensen]

Does anyone have any insight on if there's a way to 'extract' the baseline model that Lavaan is running in the background?

The fitted baseline.model is not saved anywhere, but you can fit the model that lavaan uses by default.  Just submit your fitted model to the function:

example(cfa)
baseline.parTable <- lav_partable_independence(fit)
baseline.fit <- update(fit, model = baseline.parTable)
summary(baseline.fit)

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen

[Allie Choate]

Perfect, thank you so much!