Variable selection in Poisson regression model using penalized likelihood methods
The Poisson regression model is one of the most important models of linear logarithmic regression it is the tool by which the dependent variable is modeled when the values of that variable are in the form of count values. As with other regression models, the model may contain many explanatory variables, which negatively affect the accuracy of the model and its simplicity in interpreting the results. The aim of this study is to review and compare the methods of selecting variables in the Poisson regression model through the methods of penalization using simulations and real data. The Monte-Carlo method has been used in simulations to generate data following the Poisson regression model according to factors such as sample size, simple correlation coefficient value and number of independent variables. Two aspects of the evaluation of the performance of penal methods has been based on: the first is the evaluation of the accuracy of the prediction and the second is the assessment of the choice of variables as a benchmark for comparison. In addition, the methods of penal potential were applied to real data collected from patients with chronic renal insufficiency and who are treated with continuous dialysis. Patients were diagnosed by specialized doctors in collaboration with Ibn Sina Medical Center - Synthetic College Unit.