Reproducing results, robust CFI giving NA

[Chris Y.]

Dear lavaan users

For a paper I am revising, I am trying to reproduce a model I calculated with lavaan 06-1.1208 in May 2018. In contrast to the output I got last May, the robust CFI, robust TLI and robust RMSEA now print as NA, whereas I was getting a figure for both last time. I get the NA results both if I use the current lavaan 06-3. version, as well as if I install versions 06-1 or 06-2. 

I have tried installing 0.6-1.1208 as follows, but it does not seem to be available.

install_version("lavaan", version = "0.6-1.1208", repos = "http://cran.us.r-project.org")

I add information on my model below, but this seems to be a problem of lavaan versions. The version history lists this behaviour as a bug for 0.6-1. But it seems strange that 06-1.1208 did not have this bug, but it reappears in 0.6-2 and 0.6-3-

My question  is: (1) Is there a way of installing 0.6-1.1208? and (2) Is there a different reason I'm getting NA on these robust fit indices?

Many thanks for any pointers

Chris

******************

my model:

******************

I am calculating a SEM using a dataset of n=300, using fiml to account for the missing data and MLR to account for the multivariate non-normality of the variables/residuals. My syntax (simplified) is as follows

model <-
"#latent variable
PRS=~var1.z+var2.z+var3.z
#regressions
PRS~connectedness
var1.z~biodiversity+bio.

control
var2.z~biodiversity+bio.control
restoration ~ PRS + stress + biodiversity+bio.control
"

model.fit <- sem(model=model, data=mydata, missing="fiml",estimator="MLR")
summary(model.fit,  fit.measures=T, standardized=T)

[Terrence Jorgensen]

The version history lists this behaviour as a bug for 0.6-1. But it seems strange that 06-1.1208 did not have this bug, but it reappears in 0.6-2 and 0.6-3-

They aren't NA with example data in lavaan:

example(cfa)
summary(update(fit, estimator = "MLR"), fit = T)

Try installing the latest development version (0.6-4.***)

install.packages("lavaan", repos = "http://www.da.ugent.be", type = "source")

 

Is there a different reason I'm getting NA on these robust fit indices?

If you can post enough data to reproduce the problem (assuming it still occurs in the latest software), that would help track down the issue.

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

[Chris Y.]

Hi Terrence

Thanks very much for having a look at this. I'll check using 0.6-4 and post some data if I don't get anywhere with it.

Best, Chris