Dobór zmiennych objaśniających metodą najlepszego podzbioru do modelu regresji logistycznej Firtha

The application of logistic regression to small-sized data sets results in biased estimates and often leads to a complete separation problem. Under the small sample scenario the Firth’s approach to logistic regression or its Bayesian counterpart are known to solve both issues. The main goal of this study is to explore the effectiveness of the best subset variable selection algorithm, applied to both the classical and the Bayesian logistic regression.