Events Calendar

02 Dec
Dissertation Defense: Bang Wang
Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students, Postdocs

University Unit
Department of Statistics
Subscribe
Google Calendar iCal Outlook

Dissertation Defense: Bang Wang

This is a past event.

Dissertation defense by Bang Wang titled "Identification of Causal Effect Modifiers for Time to Relapse and A Weighted Generalized Win-Odds Regression Model for Composite Endpoints".

Abstract: This dissertation addresses the general topic of treatment effect evaluation and comprises two distinct projects. In project I, we create an intuitive and readily implementable framework to facilitate the discovery of treatment effect modifiers (i.e., moderators) and to make treatment recommendations for time-to-event outcomes. To minimize the impact of a misspecified main effect and avoid complex modeling, we construct the framework by matching the treated with the controls and modeling the conditional average treatment effect via regressing the difference in the observed outcomes of a matched pair on the average moderators. Censoring is handled by the Inverse-probability-of-censoring weighting. After matching, the framework can be flexibly combined with popular variable selection and prediction methods such as linear regression, LASSO, and random forest to provide various combinations of potential moderators. The optimal combination is determined by the out-of-bag prediction error and the area under the receiver operating characteristic curve in making correct treatmentrecommendations. We compare the performance of various combined moderators through simulations and real data analysis. Our approach can be easily implemented using existing R packages, resulting in a straightforward optimal combined moderator to make treatment recommendations. The time-to-first-event analysis is frequently employed in studies involving multiple event times in which all components are treated equally regardless of their clinical significance. In project II, we focus on Win-Odds which can handle different types of outcomes and allow for a hierarchical ordering in component outcomes. A proportional Win-Odds regression model is proposed to evaluate the treatment effect on multiple outcomes while controlling for other risk factors. The model is easily interpretable as a standard logistic regression model. However, the proposed Win-Odds regression is more advanced; multiple outcomes of different types can be modeled together, and the estimating equation is constructed based on all possible and potentially dependent pairings of a treated individuals and a control ones. In addition, informative ties are carefully distinguished from those inconclusive comparisons due to censoring and the latter is handled via the IPCW. We establish the asymptotic properties of the estimated regression coefficients using the U-statistic theory and demonstrate the finite sample performance through numerical studies.

Committee Chair and Advisor: Dr. Yu Cheng

Friday, December 2 at 2:00 p.m. to 4:00 p.m.

Posvar Hall Department of Statistics, Seminar Room

Dissertation Defense: Bang Wang

Dissertation defense by Bang Wang titled "Identification of Causal Effect Modifiers for Time to Relapse and A Weighted Generalized Win-Odds Regression Model for Composite Endpoints".

Abstract: This dissertation addresses the general topic of treatment effect evaluation and comprises two distinct projects. In project I, we create an intuitive and readily implementable framework to facilitate the discovery of treatment effect modifiers (i.e., moderators) and to make treatment recommendations for time-to-event outcomes. To minimize the impact of a misspecified main effect and avoid complex modeling, we construct the framework by matching the treated with the controls and modeling the conditional average treatment effect via regressing the difference in the observed outcomes of a matched pair on the average moderators. Censoring is handled by the Inverse-probability-of-censoring weighting. After matching, the framework can be flexibly combined with popular variable selection and prediction methods such as linear regression, LASSO, and random forest to provide various combinations of potential moderators. The optimal combination is determined by the out-of-bag prediction error and the area under the receiver operating characteristic curve in making correct treatmentrecommendations. We compare the performance of various combined moderators through simulations and real data analysis. Our approach can be easily implemented using existing R packages, resulting in a straightforward optimal combined moderator to make treatment recommendations. The time-to-first-event analysis is frequently employed in studies involving multiple event times in which all components are treated equally regardless of their clinical significance. In project II, we focus on Win-Odds which can handle different types of outcomes and allow for a hierarchical ordering in component outcomes. A proportional Win-Odds regression model is proposed to evaluate the treatment effect on multiple outcomes while controlling for other risk factors. The model is easily interpretable as a standard logistic regression model. However, the proposed Win-Odds regression is more advanced; multiple outcomes of different types can be modeled together, and the estimating equation is constructed based on all possible and potentially dependent pairings of a treated individuals and a control ones. In addition, informative ties are carefully distinguished from those inconclusive comparisons due to censoring and the latter is handled via the IPCW. We establish the asymptotic properties of the estimated regression coefficients using the U-statistic theory and demonstrate the finite sample performance through numerical studies.

Committee Chair and Advisor: Dr. Yu Cheng

Friday, December 2 at 2:00 p.m. to 4:00 p.m.

Posvar Hall Department of Statistics, Seminar Room

Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students, Postdocs

University Unit
Department of Statistics

Powered by the Localist Community Events Calendar ©