Abstract
Latent class analysis in two-way contingency tables usually suffers from unidentifiability problems. These an be overcome by using Bayesian techniques in which prior distributions are assumed on the latent parameters. Application of these techniques, however, involves some unusual computational difficulties. Bayesian techniques suitable for latent class analyses together with a remedy to the computational difficulties are discussed in this paper.
Original language | English |
---|---|
Pages (from-to) | 557-563 |
Number of pages | 7 |
Journal | Biometrika |
Volume | 76 |
Issue number | 3 |
DOIs | |
State | Published - Sep 1989 |
Keywords
- Adaptive importance sampling
- Bayesian estimation
- Contingency table
- Dirichlet prior
- Latent class model