Graduate Students Seminar
Online on Blackboard Collaborate
|Session Chair:||Neha Agarwala|
|Discussant:||Dr. Yi Huang and Dr. Weining Kang|
Speaker 1: Jing Wang
- Extending Inferences from a Randomized Trial to a Target Population: a Potential Outcome Perspective
- Study validity can be divided into internal validity and external validity. In recent years, a lot more attention has been paid to increase external validly, especially for randomized controlled trials (RCTs), which is considered as internally valid due to randomization.
- According to specification of the target population, external validity of RCTs can be further distinguished between two possible problems: generalizability and transportability. The term “generalizability” is used when the target population coincides, or is a subset of, the trial-eligible population, while “transportability” is used when the target population includes individuals who are not trial-eligible. To identify the potential outcome means in the target population from the observed trial result, a set of assumptions need to be made: 1) consistency of potential outcomes, 2) mean exchangeability in the trial, 3) positivity of treatment assignment, 4) mean exchangeability in the target population for trial participants and non-participants, 5) positivity of trial participation. Given those assumptions above, the marginal potential outcome mean in the target population can be estimated by outcome model-based estimators, IP weighting estimators, and doubly-robust estimators (augmented IP weighting estimators).
Speaker 2: Daniel Kelly
- Comparing Time and Accuracy for Missile Trajectory Numerical Methods
- The computation of the error for a missile trajectory involves the use of many different numerical analysis methods. The study explains how each method is computed and compares each method’s performance to conclude which is the most efficient method for large scale error calculations. The study runs a simple random trajectory against each method for a varying number of points with step size changes and compares the accuracy and time each method took. The presentation will explain the numerical analysis behind calculating the trajectory error as well as present the observed results with the summary of the most efficient method.