Overview
Regression model selection is not a matter of trying everything and picking the highest R². The right model is determined by the data-generating process, the relationship structure between predictors and outcome, the sample size relative to feature count, and the interpretability requirements of the use case. A neural network that achieves 0.92 R² on training data but 0.61 on holdout is not a better model than a regularized linear regression that achieves 0.78 on both.
Most model selection processes skip the assumption checks that determine which models are valid, apply regularization without understanding what it does to the coefficient estimates, and evaluate performance on metrics that don't match the business objective.
The Regression Model Selection & Specification Prompt generates a complete regression modeling specification: assumption diagnostics, model selection logic by data structure and objective, regularization parameter selection, and a validation framework that tests genuine out-of-sample predictive power.
What you get: - Assumption diagnostic protocol for linear and nonlinear regression - Model selection matrix by data structure and objective - Regularization type and strength selection - Hyperparameter tuning specification - Out-of-sample validation framework
Built for: data scientists and analysts who need defensible, assumption-checked regression models for prediction and inference.