Estimation and testing in large binary contingency tables

Very sparse contingency tables with a multiplicative structure are studied. The number of unspecified parameters and the number of cells are growing with the number of observations. Consistency and asymptotic normality of natural estimators are established. Also uniform convergence of the estimators to the parameters is investigated, and an application to the construction of confidence intervals is presented. Further, a family of goodness-of-fit tests is proposed for testing multiplicativity. It is shown that the test statistics are asymptotically normal. The results can be applied in such different fields as production testing or psychometrics.