Comparison of the method of partial least squares and the algorithm of singular values decomposion to estimate the parameters of the logistic regression model in the case of the problem of linear multiplicity by using the simulation

Abstract

The logistic regression model is an important statistical model showing the relationship between the binary variable and the explanatory variables. The large number of explanations that are usually used to illustrate the response led to the emergence of the problem of linear multiplicity between the explanatory variables that make estimating the parameters of the model not accurate. The methods used to estimate the parameters of the logistic regression model in the case of the linear multiplication problem. These methods are the method of regression of the partial least squares and the algorithm of singular value decomposion. The simulation method was used to compare estimation methods through the mean error squares of the model. It has been shown through the comparison that the algorithm of singular value decomposion is best in estimating the parameters of the logistic regression model in the case of the problem of linear multiplicity