How Big a Sample Do I Need?
We do not know of simple ways to estimate sample size needs or power for LCA or LTA. Such estimates may require complex Monte Carlo studies. We have been able to conduct conditional RMLCA with 27 repeats of a binary indicator with covariates in samples as small as 200 subjects, but treated this and other small (N ranging from 287 to 407) datasets as replications of an analysis conducted in a larger (N=1433) sample and encountered some estimation problems in conditional models using small samples.