A product-multinomial framework for categorical data analysis with missing responses

摘要: We extend the multinomial modeling scenario for analysis of categorical data with missing responses described by Paulino (1991, Brazilian Journal Probability and Statistics, 5, 1-42) to product-multinomial setup so that inclusion explanatory variables is allowed. Assuming an ignorable mechanism, linear log-linear models may be fitted via maximum likelihood. Weighted least squares methodology as well used fit more general functional models, if a completely at random mechanism assumed. also consider hybrid approach, where any missingness process likelihood in first step, estimated marginal probabilities categorization their covariance matrix are second stage model weighted squares, spirit asymptotic regression methodology. Goodness-of-fit tests present, illustrated two sets. All methods were computationally implemented subroutines written R.