In most cases the statistical association does not perfectly fit the deterministic models of logical or numerical covariation just described; that is, there is variation from the model. Tests of statistical significance are required to measure the degree to which data fit or vary from one of these models. Formal measures of statistical covariation depend on the type of variation of the measures of each
variable involved: x^2 tests may be used to judge the significance of the association between categorical variables, and t-tests or analyses of variance are used to judge the significance of mean values of an interval variable across groupings of a categorical variable.