Achieving expected parities: a reanalysis of Freedman et al.'s data, 1962-1977.

A reanalysis of data included in a recent report by Freedman et al. (1980), using methods suggested by Goodman (1978), yields considerably different substantive conclusions than were arrived at by those earlier analysts. These differences clearly point out the hazards of descriptive analysis, especially of sample data which come in the form of contingency tables.