regression
Ridge regression
Fit a linear regression when predictors are correlated and coefficient shrinkage is useful.
Requirements
- continuous outcome
- predictor(s)
- penalty alpha
Avoid when
- sparse feature selection is the goal
- unpenalized inference with p-values is required
Formula
arg minβ ||y−Xβ||² + λ||β||²₂
Interpretation
Coefficients are shrunk toward zero; p-values are not reported.
Worked example
Predict a continuous outcome from many correlated biomarker predictors.
Open example datasets