I would like to do a Latent Class Analysis on the topic of poisoning deaths. I have data on substances in the blood, F diagnoses, gender, age... The problem is that I don't know how to group the substances into a reasonable number of groups. For example, fentanyl, should it be separately or with all opioids. Should Z drugs be included separately or with benzodiazepines. The recommendation is not to include substances that are less than 5% in the sample. Most of these substances when you put them in the sample are from 10 to 20%. On the other hand, fentanyl is very fatal on its own. You don't need to combine it with other substances a lot to cause death. If anyone has any advice, feel free to write here. In addition, should I put sex, age, F diagnosis in the LCA or as covariates in further analysis.