Hello, I am currently working on my dissertation proposal and one of my committee members asked me about reliability testing for Latent Class Analysis. Another committee member suggested that I repeatedly and randomly select 75% of my sample (n = 290) and run the LCA to see if I get the same number of classes. From my understanding, I thought that BLRT creates multiple bootstrap samples to arrive at the significance of the k class in comparison to k-1 class. Therefore, it sounds to me like BLRT would be better at answering the reliability of the classes question than running multiple LCA's with 75% of my sample. However, I am having a hard time finding the research to better understand BLRT. I would greatly appreciate any feedback on my belief that BLRT would meet this committee members request regarding reliability, rather than having to run multiple LCA's with 75% of the sample.
It is unclear what reliability should mean for LCA. Perhaps classification quality and entropy is what is meant. I haven't seen 75% repeated sampling used. I don't view BLRT as serving the same function because the 75% idea does not test k-1 versus k classes. BLRT is described in the Nylund et al paper and its references.