|
Abstract:
|
There is increasing interest in classifying disease on the basis of a moderate number of genetic factors. For binary outcomes, latent class analysis (LCA) is a natural approach. However, LCA is complicated when the number of subjects is of small to moderate size. We present a regularization approach, penalized LCA, that improves estimation. We consider an LCA model where, conditional on unobserved membership in one of k classes, item responses are independent. Using a family of penalties, including the ridge and lasso, we extend this model to address higher-dimensional problems. Penalization may be imposed across items when the responses have a natural ordering, or may be imposed across classes when it is of interest to identify items that distinguish class membership. We illustrate our methods on loss-of-heterozygosity data from brain tumors and hypermethylation data in bladder tumors.
|
