A Categorical Data Analysis of Heart Disease Levels Using Generalized Linear Models and Odds Ratios

Abstract

The analysis of risk factors affecting the probability of category levels due to heart disease is evaluated. Two cases are considered, case 1: responses are divided into two groups: death and survivors, case 2: responses are divided into 4 groups: non-heart disease and living, heart disease but still living, non-heart disease but death, and heart disease and death. The medical risk factors consist of age, body mass index, blood pressure levels, cholesterol levels, stress levels. The research real data with 200 patients were from Afifi and Azen (1979). All analyses are performed using the probit model, logit model and baseline-category logit model run with SAS Enterprise Guide 5.1. The results show the risk factors under the probit model and the logit model are age, blood pressure levels, cholesterol levels and the diagnosis of cardiac ischemia, significantly at 0.05. Those factors from the baseline–category logit models are also age and blood pressure levels, significantly at 0.05. The three generalized linear models are all statistically adequate of fits for the data, significantly at 0.05. It is also found from odds ratio such that the risk factors for age is 3.21266 under the logit model.