Transporting a prediction model to a new cohort with a novel resampling method

Principal Investigator

Name
Lei Song

Degrees
PhD

Institution
George Washington University

Position Title
student

Email
leisong@gwu.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-1182

Initial CDAS Request Approval
Jan 9, 2024

Title
Transporting a prediction model to a new cohort with a novel resampling method

Summary
When building a prediction model, researchers are often interested in applying it to a new cohort. But it is very likely that the new target population has a potentially different distribution compared with the source population where the model is developed, and reweighting is a popular approach when such covariate shift occurs. But many reweighting methods suffers from unstable performance when the population overlap is not sufficient or when under the present of extreme weights. Thus a robust resampling method is develop to help improve the prediction accuracy to transport a predicting model. It is of our interest to test how it will perform under different degree of overlap in NLST dataset, compared with other existing approaches.

Aims

1. Measure the difference among dataset from different centers.
2. Develop and apply the novel reweighting method to the NLST dataset.
3. Compare the model accuracy with baseline model and other existing transporting approaches.

Collaborators

Dr. Qing Pan from George Washington University
Dr. Chen Hu from Johns Hopkins university