Estimating inverse-probability weights for longitudinal data with dropout or truncation: The xtrccipw command.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Statistical Methods

Background:

• Longitudinal studies often experience participant dropout, leading to unobserved outcomes.
• Outcome data can be truncated (e.g., by death), rendering further observations meaningless.
• Existing methods like regression conditioning on continuation (RCC) handle missingness but require specialized implementation.

Purpose of the Study:

• To introduce the `xtrccipw` command for implementing regression conditioning on continuation (RCC).
• To provide a user-friendly tool for estimating inverse-probability weights (IPWs) and applying them within generalized estimating equations (GEE).
• To facilitate analysis of longitudinal data with monotonically missing outcomes, even in the presence of truncation.

Main Methods:

• The `xtrccipw` command estimates inverse-probability weights (IPWs) for individuals continuing in the study.
• It then applies these IPWs within a generalized estimating equations (GEE) framework, leveraging the `glm` command.
• The command can also perform weighted GEE analysis when truncation is absent.

Main Results:

• The `xtrccipw` command successfully estimates IPWs and fits RCC models.
• Demonstrations using example and original Kurland and Heagerty (2005) data validate the command's functionality.
• A simulation study illustrates the empirical properties of the RCC method.

Conclusions:

• The `xtrccipw` command offers a practical solution for analyzing longitudinal data with missing outcomes.
• It effectively implements regression conditioning on continuation (RCC) using inverse-probability weights (IPWs).
• The command enhances the analysis of studies affected by dropout and potential truncation.