regularized logistic regression - andrew gelman

priors11.pdf (Objekt application/pdf)

cf problems of logistic model overfitting /cf iterative optimiztion /, data sparsity, complete separation, multicolinearity in LOGISITIC REGRESSION
- one solution - exclude variables that cause complete separation, multicolinearity etc. - however these variables may be the strongest, and of importance.
- new solution - adding some data into model /priors/, which may improve/stabilize model performance

technical