Degree Type

Creative Component

Semester of Graduation

Summer 2019

Department

Statistics

First Major Professor

Ulrike Genschel

Degree(s)

Master of Science (MS)

Major(s)

Statistics

Abstract

There were 20.39 million military veterans in the U.S. in 2016 and more than 65.5% of them being 55 years or older. According to T. Simpson et al. (2012) and Lehavot et al. (2012), veterans had worse overall health, a higher incidence of health risk behavior and chronic health conditions than civilians (non-veterans). However, the causal relationship between the military experience as a veteran and physical health cannot be determined. We applied the Rubin Causal Model on the BRFSS2016 data to find if there is any significant difference between the veterans and the non-veterans in Heart Attack incidence rate. We also used the Logistic Regression Model to understand the effect of being a veteran and predict the probability of each individual having had Heart Attack(s). We first used categorical variables to replace all numerical variables and applied an exact 1-to-1 matching. Then, we applied another exact 1-to-1 matching on the key numerical variable (Age). After the matching, the Heart Attack incidence rates were 9.94% in the veterans and 9.23% in the non-veterans. The difference was 0.71% and the p-value was 0.00035. In the logistic regression model(s), the effect of being a veteran was positive (increasing the probability to have had a heart attack) and significant. Also, we show that when the data size reduces the simple model could predict better than the complex model.

Copyright Owner

Yuchen Wang

File Format

PDF

Share

COinS