Log-linear models are a type of statistical model used to describe the relationship between categorical variables by modeling the logarithm of expected frequencies in contingency tables. These models are particularly useful for analyzing multi-way tables and understanding the interaction effects among variables without assuming a specific causal direction.