ABSTRACT
Incomplete 2 × 2 tables are often encountered in statistical analysis. It is clear that all the aforementioned approaches aim at inference, based on an identifiable sampling distribution that is acceptable to all, no matter which untestable additional assumptions an analyst may or may not introduce in order to resolve the identification issue. It is important to distinguish between the study of identification, provided an infinite amount of data under the given structure, and statistical inference from finite samples. There are several related approaches within the matched data setting. In ecological inference, the observed data are the margins of the unobserved complete 2 × 2 table. In ecological inference, the observed data are the margins of the unobserved complete 2 × 2 table.
