Multinomial Logistic Regression Analysis of Smoking Status and Demographic Factors in Malaysia

Abstract

This study investigates the association between demographic factors and smoking behavior among individuals in Malaysia, utilizing a secondary dataset of 643 subjects categorized as daily smokers, ex-smokers, and non-smokers. The analysis focuses on eight independent variables: weight, height, gender, ethnicity, marital status, education level, occupation, and physical activity. Descriptive statistics are employed to summarize the characteristics of the sample, while multinomial logistic regression is employed to identify significant predictors of smoking status. The results indicate that the final model, which includes predictors, significantly improves the fit compared to the intercept-only model (p-value < 0.05). This study identified height, gender, ethnicity, and education level as significant predictors of daily smoking behavior. Males and individuals from the Malay ethnic group show higher odds of being daily smokers, while higher education was associated with lower smoking rates. However, the goodness of fit test (p-value <0.001), indicates that the model does not fit the data well. This difference suggests that while the model captures some important relationships among variables, it fails to function as a predictive model for smoking behavior.