摘要：The practical analyses of interactions between categorical variables in various areas (such as public opinion researchor marketing research) are often only applications of chi-square tests in two-way contingency tables. However,in many situations it is impossible to use large-sample approximations to sampling distributions when theiradequacy can be in doubt. It is known, that these approximations may be very poor when the contingency tablecontains very small expected frequencies. However, recent work has shown that these approximations can bevery poor when the contingency table contains both small and large expected frequencies. Of course, the rule ofthumb of a minimum expected frequency is not met either in the case of sparse table. The article deals with alternativeapproaches to the data analysis in such cases. It points out other possibilities and shows that thanks to thedevelopment of computer technology exact methods previously only difficult usable are available for this purpose.
关键词：Contingency tables; categorical data analysis; exact inference about associations
