|
Activity Number:
|
37
- Novel Semiparametric Methods for Causal Inference
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, August 7, 2022 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
SSC (Statistical Society of Canada)
|
|
Abstract #319243
|
|
|
Title:
|
A Semiparametric G-Computation Approach Based on Cumulative Probability Models
|
|
Author(s):
|
Andrew J Spieker* and Caroline Birdrow and Bryan Shepherd
|
|
Companies:
|
Vanderbilt University Medical Center and Vanderbilt University Medical Center and Vanderbilt University Medical Center
|
|
Keywords:
|
causal inference;
semi-parametric;
g-compuation;
cumulative probability models;
cost
|
|
Abstract:
|
Standardization is a well-known method to account for confounding in settings where the exposure of interest is stable over time. Nonparametric Monte Carlo-based standardization can be conducted in this setting based on the empirical distribution of baseline confounders. G-computation is a longitudinal generalization of standardization suitable for settings in which there is time-dependent confounding. While g-computation is a useful tool for estimating longitudinal causal effects, its reliance on parametric models is sometimes criticized. Recently, cumulative probability models have been proposed as a semi-parametric approach to analyzing continuous outcomes. Through this approach, one is able to recover the cumulative distribution function of an outcome conditional on covariates. In this talk, we discuss the utility of cumulative probability models for use in g-computation as a way to avoid overly stringent parametric assumptions. We illustrate the utility of this methodology through a range of simulation studies and an application to a large cohort study of women with endometrial cancer to compare cumulative medical costs associated with various adjuvant treatment strategies.
|
Authors who are presenting talks have a * after their name.
