Good evening, I am considering running a LPA using 7 scales (each with apprx. 5 items). In the literature I've reviewed I've seen examples where others have used the scale scores, rather than the individual items as class indicators, but none have provided the rationale for doing so. I am hoping someone here might be willing and able to help.
I ran a latent class growth mixture model on certain outcome variables over time before realizing that these variables needed to be scaled to match a comparative outcome variable. To scale the variables, I divided by the range of the measure. When I reran the model, the number of classes found to be significant changed (6 instead of 3). Why would scaling the outcome variable affect the number of classes? Why wouldn't the slopes for the original classes be the only thing affected?