Identifying Variable Relations Using Relevance Logic

In data analysis that is based on methods from the generalized linear model, variables are related to each other, and predictions are made in which variables are linked. Issues of discussion concern strength of relation and direction dependence. Relations are of the regression type. In this article, the discussion focuses on type of relation. Based on relevance logic, conjunctive, disjunctive and implicative relations are discussed, in particular with respect to their applicability in data analysis. It is illustrated how these relations can be distinguished from each other in cross-classifications of empirical data. It is also shown how predictions can be made of type of relation instead of variables themselves. Empirical data examples illustrate these predictions. Possible extensions are discussed in which relations of the type 'if and only if' are used.