Revue internationale des sciences appliquées - Recherche et examen Libre accès

Abstrait

Assessing Discrimination Power for Binary Logistic Regression Model Based on Parametric and Non-Parametric Methods

Md. Asadullah

The main focus of this paper is to measure the discrimination ability of the fitted binary logistic regression model after admission the patients in ICU (intensive care unit). In this paper we use parametric and non-parametric methods for measuring discrimination ability of the logistic regression classifier. The most important analysis in which the outcome variable is binary or dichotomous. It can be used to predict a binary dependent variable from a set of independent variables. Since our outcome variables have binary categories, so binary logistic regression prefers to estimate model parameters. This technique is preferred by many researchers in the analytical fields. It is also widely used in various clinical researches to predict the risk of a patient's future health status. Predictions based on these models have an important role in predicting the survival of patients in ICU. Concordance statistic (C-statistic), which is equivalent to the area under a receiver operating characteristic curve (AUC), is frequently used to quantify the discriminatory power of the logistic model because of its straightforward clinical interpretation. In this paper we assess the discrimination power in simulation and real data for binary logistic regression