TIME-VARYING EFFECT MODEL FOR STUDYING GROUP DIFFERENCE IN HEALTH BEHAVIOR

Restricted (Penn State Only)
Author:
Yang, Songshan
Graduate Program:
Statistics
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
November 16, 2017
Committee Members:
  • Runze Li, Thesis Advisor
  • Bing Li, Committee Member
  • Matthew Logan Reimherr, Committee Member
Keywords:
  • Time varying-coefficient effect model
Abstract:
The thesis includes two projects related to the detection of group difference of health risk behavior using time-varying effect models. In the first project, we proposed to use a time-varying effect model to characterize gender-specific trajectories of health behaviors, and we also conducted hypothesis testing for gender differences. We first used two motivating examples which demonstrated that the proposed model works for both the longitudinal studies and the short-term studies that involve intensive data collection. Then a simulation study is conducted, which shows that the accuracy of estimation of trajectory functions improves as the sample size and the number of time points increase. In terms of the performance of the hypothesis testing, the type I error rates are close to their corresponding significance levels under all combinations of sample size and number of time points. In addition, the power increases as the alternative hypothesis deviates more from the null hypothesis, and the rate of this increasing trend is higher when the sample size and the number of time points are larger. The second project proposed to use a time-varying effect model for examining group differences in trajectories of zero-inflated count outcomes. The motivating example demonstrates that this zero-inflated Poisson model allows investigators to study group differences in different aspects of substance use (e.g. the probability of abstinence and the quantity of alcohol use) simultaneously. The simulation study shows that the accuracy of estimation of trajectory functions improves as the sample size increases; the accuracy under equal group sizes is only higher when the sample size is small (100). By examining the performance of the hypothesis testing, we found that the type I error rates are close to their corresponding significance levels under all settings. On the other hand, the power also increases as the alternative hypothesis deviates more from the null hypothesis and the rate of this increasing trend is higher when the sample size is larger. Moreover, the hypothesis test for the group difference in the zero component tends to be less powerful than the test for the group difference in the Poisson component.