Suitability of Grade-of-Membership Techniques to Correct for Selection Bias inthe Social Health Maintenance Organization Evaluation

摘要

The purpose of the study is to assess the suitability of grade of membership(GoM) analysis to correct for selection bias in the social health maintenance organization demonstration. The researchers discuss GoM in the context of data reduction techniques. Following in the path of latent structure analysis, GoM seeks to illuminate the underlying communalities that relate the variables measured. The researchers conclude that GoM does not correct for selection bias, or bias based on unobserved variables. Comparison of different GoM specifications can inform one as to the presence of selection bias. The researchers suggest testing for the presence of selection bias by estimation of a selection-corrected outcome equation. The researchers also provide several recommendations related to the use of GoM to deal with missing data problems.