Latent Class Analysis--classification error and residual correlation?

[Gracie Venechuk]

Hi all,

I'm using LCA with distal regression in a paper and one of my reviewers asked about classification error and residual correlation among my indicator items. From what I can see, these are statistics you can usually generate using MPLus and Latent Gold, but there doesn't seem to be a way to do this naturally in Stata. Some additional notes:

1. Entropy for my model is quite good--.92--so I feel fairly confident about classification. BUT I know that reporting entropy alone may not be satisfactory
2. I'm aware of the Penn State BCH plug in that corrects for classification error; however, I'm a little wary of using it given that it's no longer maintained

If anyone has had similar experience and has suggestions, hacks, advice etc., I'd be most grateful.
Thanks in advance for your time and your feedback!

[Roman Mostazir]

Please see the FAQ section for posting rules in the forum. You are expected to provide the Stata codes you used for your analysis and post the codes using code delimiters. Otherwise it is difficult to understand what you are trying to achieve. For example, you mentioned about 'residual correlation' which, I understand is covariance of error terms and not necessarily a statistic as you had assumed. Fitting LCA with -gsem- error terms can be asked to covary.