Handling Overdispersion Problems in Multinomial Logistic Regression (Study Case in Stress Level Data)

Abstract

The development of statistical methods also impacts the development of analytical methods. One analytical method in which this is the case is the multinomial logistic regression modeling method. In this method, we have more than two categories of the response variable. At this time, the data used in modeling has various problems, one of which is overdispersion. This is a condition where there is a correlation between the response variables. This paper will examine the performance of multinomial logistic regression when there is overdispersion present in the data. We will focus on implementing methods in the Stress Level Data, which is about student stress level due to ‘zoom fatigue’. The model selection is carried out using the stepwise method, where the best model is selected based on the smallest AIC value of the model candidates. The best model for our data shows that the performance of the multinomial logistic regression approach with overdispersion treatment is better than without allowing for overdispersion.