← Back to Modules Directory

Module 10 - Link Functions

Generalized Linear Models (GLMs) extend the capabilities of traditional ordinary least squares architectures to accommodate non-normal, skewed, or discrete target variables. A foundational component of the GLM framework is the link function, which establishes a strict mathematical connection between the un-bounded linear predictor space and the bounded expected value of the target distribution.

Common Link Functions and Target Mappings

Canonical Selection Criteria

Isolating the appropriate link function parameter configuration depends on three architecture rules:

Selecting an inappropriate link function model specification introduces structural estimation bias and renders downstream hypothesis inferences invalid.