Tracking, Weighting, and Sample Selection Modeling to Correct for Attrition
Published in: Proceedings of the Section on Survey Research Methods, American Statistical Association, v. 1, 1995, p. 402-407
Posted on RAND.org on January 01, 1995
The authors' validation analyses show excellent agreement between weighted estimates under simulated attrition and the true values observed at baseline, and good adjustments at the three-year follow-up. While tracking has good face validity, it would not have been an adequate substitute for statistical adjustment to correct for attrition. The results support the importance of evaluating the effects of assumptions underlying any approach to nonresponse adjustment.