Efficient direct sampling MCEM algorithm for latent variable models with binary responses

摘要: While latent variable models have been successfully applied in many fields and underpin various modeling techniques, their ability to incorporate categorical responses is hindered due the lack of accurate efficient estimation methods. Approximation procedures, such as penalized quasi-likelihood, are computationally efficient, but resulting estimators can be seriously biased for binary responses. Gauss-Hermite quadrature Markov Chain Monte Carlo (MCMC) integration based methods yield more estimation, they much intensive. Estimation that achieve both computational efficiency accuracy still under development. This paper proposes an direct sampling EM algorithm (DSMCEM) with Mixed effects item factor analysis used illustrate this algorithm. Results from two simulation studies a real data example suggest that, compared MCMC EM, DSMCEM significantly improve well produce equally parameter estimates. Other aspects extensions discussed.