Abstract: The analysis of incomplete contingency tables is an interesting problem, which is also of practical interest. In this talk, we propose various missing data models for analyzing arbitrary three-way and multidimensional incomplete tables, and study estimation and testing under them. We also explore the problem of boundary solutions in such tables, proposing their various forms, connections among them and establishing sufficient and necessary conditions for their occurrence (using only the observed cell counts), which prove useful for model selection. Finally, we suggest methods for assessment of the missing data mechanisms of variables in the above tables in terms of only the observed cell counts.