Lavaan CFA equivalent of CMIN/DF (AMOS)

[Scott Wagstaff]

My university course uses AMOS, but being a renegade I decided to use Lavaan instead.

One of the acceptance criteria (supplied automatically by AMOS) stipulated by my teacher is that CMIN/DF should be <2 (but 2-5 is acceptable) where CMIN is Chi Square statistic and DF are the degrees of freedom

Am I correct in thinking that the cfa function in lavaan which produces the value for "Minimum Function Test Statistic" is equivalent to CMIN in AMOS, and all I need to do is divide this figure by the degrees of freedom to get CMIN/DF and make my teacher happy?

From the output below I think my CMIN/DF value is 118.367 / 98 = 1.208

If my assumptions are correct, is there a way to get lavaan to generate this number automatically or should I just present the manual calculation?

I have also noted that to align with the results my teacheris expecting from AMOS I should probably use the Wishart function when performing the cfa - any comments if this will be the correct approach?

## lavaan (0.5-23.1097) converged normally after 44 iterations
##
## Number of observations 400
##
## Estimator ML
## Minimum Function Test Statistic 118.367
## Degrees of freedom 98
## P-value (Chi-square) 0.079

[Terrence Jorgensen]

Am I correct in thinking that the cfa function in lavaan which produces the value for "Minimum Function Test Statistic" is equivalent to CMIN in AMOS, and all I need to do is divide this figure by the degrees of freedom to get CMIN/DF and make my teacher happy?

Yes, they are talking about the chi-squared to df ratio, which is one of the very first proposed fit indices to accompany a test of significance of model misfit.  It has severe limitations (as do all indices, but later ones fix some problems associated with this one).  But yes, you just need to calculate this ratio:

myFits <- fitMeasures(fit)
myFits["chisq"] / myFits["df"]

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

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

UvA web page: http://www.uva.nl/profile/t.d.jorgensen

[Scott Wagstaff]

Thanks for the help and quick feedback!

Scott