Inconsistency in LMR Test in tidyLPA compared to Mplus

[Stiltner, Brendan]

Hello,

 

I am working on a project using latent class analysis in substance use disorders. While I began the project using R and your package to calculate the Lo-Mendell-Rubin Test, I’ve began to try to use Mplus, as it seems to be the standard for LCA in this area. However, when calculating the LMR Test with the same data in tidyLPA in R and Mplus, I get different results and I am not sure why.

I am wondering if it may be because the LMR Test are differently implemented or calculated between latent profile analysis and latent class analysis. If so, because the variables I am using are binary rather than continuous, and your package is designed to calculate the LMR test for LPA rather than LCA, might this explain why my results are different between tidyLPA and Mplus?

If this should not cause a difference, do you know what may be causing this discrepancy? I can provide further information if needed. Thank you in advance.

 

Brendan Stiltner (he/him)

Postgraduate Associate

Polimanti Lab (Division of Human Genetics)

Psychiatry Department, Yale School of Medicine