Bayesian Analysis Of The Ordinal Categorical Data With Misclassification

Misclassification happened when a subject has been classified into a category that does not reflect the subject's state. In general, double sampling technique is useful to analyze the misclassified categorical data. Based on the latent normal variable models, we develop a Bayesian method to analyze multivariate ordinal categorical data with misclassification. Gibbs sampler and Metropolis-Hastings algorithm are used to generate samples from the posterior of the parameters. To obtain a rapidly convergent algorithm, the parameter expansion technique is applied to the correlation structure of multivariate model. The proposed model and sampling method are demonstrated by simulation and are applied to analyze the real data.