Abstract
The estimation of binary responses in factor analysis models is often complicated, because the marginal likelihood involves an intractable integral. When the number of latent variables is large, the dimensionality of a required integral will be high, and thus numerical integration would not be an ideal estimation method. This paper proposes H-likelihood for the estimation of binary response factor analysis models, avoiding the intractable integral. Examples and simulation studies demonstrate the performance of the proposed method.
Original language | English |
---|---|
Pages (from-to) | 72-79 |
Number of pages | 8 |
Journal | Journal of Multivariate Analysis |
Volume | 106 |
DOIs | |
Publication status | Published - Apr 2012 |
Keywords
- Binary response
- Factor analysis
- H-likelihood
- Marginal likelihood