COMPOSITE LINEAR-MODELS FOR INCOMPLETE MULTINOMIAL DATA

A composite linear model (CLM) is a matrix model for incomplete multin omial data. A CLM provides a unified approach for maximum likelihood i nference which is applicable to a wide variety of problems involving i ncomplete multinomial data. By formulating a model as a CLM, one can s implify computation of maximum likelihood estimates and asymptotic sta ndard errors. As an example, we use CLM to test marginal homogeneity f or ordered categories, subject to both ignorable and non-ignorable mis sing-data mechanisms.