Binary Logistic Regression Procedure with an Application

Abstract

Binary logistic regression is widely used in health, social, and science sciences, particularly in studies involving extensive categorical independent variables. Considering that binary logistic regression calculates the probability of a specific event occurring or not occurring relative to the opposite event using one or more independent variables, its widespread use becomes even clearer. Another reason for its widespread use is that regression equations based on the least squares method cannot be used in data sets with numerous categorical variables affecting a two-level dependent variable. Another advantage is its ability to make future predictions and identify risk factors and impact magnitudes that influence the occurrence of events. However, in practice, many errors can occur, especially during model construction. Determining the individual effects of variables, joint effects, interaction effects, and how independent variables enter the model are highly sensitive processes. This study will examine the operations performed on independent variables in binary logistic regression until the final model is formed, using a numerical example. Thus, an important resource will be created for researchers to obtain the correct models with the correct process flow in binary logistic regression.