Obtain fit measures for a model estimated with lavaan.survey

[Joshua Rosenberg]

Hi all, I have a question about Daniel Oberski's excellent lavaan.survey package; excuse me if this is off-topic for this forum.

If I fit a model, e.g.:

library(lavaan)
model <- '
 # latent variable definitions
 ind60 =~ x1 + x2 + x3
 dem60 =~ y1 + a*y2 + b*y3 + c*y4
 dem65 =~ y5 + a*y6 + b*y7 + c*y8
'

fit <- cfa(model, data=PoliticalDemocracy)
fit.measures(fit)

Then fit measures for that fitted model will be returned.

But, if I add the following lines to adjust standard errors (to account for the grouping of observations using a hypothetical grouping variable), the same fit measures will be returned, whereas the output from the `summary()` method prints both the fit statistics for the original model and that with the adjusted standard errors.

library(lavaan.survey)group <- svydesign(ids = ~group, data = PoliticalDemocracy)
fit2 <- lavaan.survey(fit, group)

fit.measures(fit2)

I checked the arguments available for `fit.measures()`, and there doesn't seem to be a way to request those from the model with the adjusted standard errors. Is there any way to obtain the fit measures from this adjusted model?

[Terrence Jorgensen]

Did you know lavaan now offers cluster-robust SEs and tests?

fit <- cfa(model, data=PoliticalDemocracy, cluster = group)
fit.measures(fit)

Is the problem that the function is called fitMeasures(), not fit.measures()?

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

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

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

[Joshua Rosenberg]

Hi, thank you, I did not know about the cluster argument - that's very helpful!

The fit measures obtained from `fit.measures(fit)` seem to be associated with the normal/not cluster-robust SEs and tests. Do you know any way around this?

[Terrence Jorgensen]

The fit measures obtained from `fit.measures(fit)`

Again, that is not a function name.  It returns an error for me, so if it works for you, you must be using a function from another package or that you have designed yourself.  The lavaan package's function is fitMeasures().

seem to be associated with the normal/not cluster-robust SEs and tests. Do you know any way around this?

I don't understand what you mean.  If you request cluster-robust SEs and tests, that is what you get.  

model <- '
    f =~ y1 + y2 + y3
    f ~ x1 + x2 + x3
'
fit <- sem(model, data = Demo.twolevel, cluster = "cluster")
fit.measures(fit) # returns an error
fitMeasures(fit)  # returns both naive and scaled tests, and indices using both
summary(fit, fit = TRUE)

The summary() output includes the naive and scaled tests/indices in separate columns.