[email protected] writes: >A friend asked me for examples of case studies that imputed missing data, >and then afterwards collected the data and validated (or invalidated) the >imputation model. (He is doing a study of that sort and would like to >cite other examples.) This is honest missing data, not simulations in >which data were artificially made missing to check model fit.
It depends on what you mean by "real missing data". We had a monthly survey where preliminary results were published before all the data was in. We tested various imputation schemes and compared results with the eventual reported values at the final deadline. It was presented by myself and Paula Weir at the 1992 Joint Meetings, and should be in the corresponding Proceedings.
