Topic: Purpose-Driven Model Specification

Worked Specification of a Forecasting Model

Worked Specification of an Explanatory Wage Model

Why Univariate Screening Is a Poor Specification Rule

Trace Candidate Predictors to the Measurement Process

Turn a Research Question Into Candidate Model Terms

Specify Predictor Availability for Each Forecast Horizon

Spot Circular and Outcome-Driven Specification Choices

Separate Identifiers from Generalizable Predictors

Retain Required Terms Even When Individual Tests Are Weak

Represent Competing Scientific Accounts With Candidate Models