NOT ASKED AND NOT ANSWERED - MULTIPLE IMPUTATION FOR MULTIPLE SURVEYS

We present a method of analyzing a series of independent cross-section al surveys in which some questions are not answered in some surveys an d some respondents do not answer some of the questions posed. The meth od is also applicable to a single survey in which different questions are asked or different sampling methods are used in different strata o r clusters. Our method involves multiply imputing the missing items an d questions by adding to existing methods of imputation designed for s ingle surveys a hierarchical regression model that allows covariates a t the individual and survey levels. Information from survey weights is exploited by including in the analysis the variables on which the wei ghts were based, and then reweighting individual responses (observed a nd imputed) to estimate population quantities. We also develop diagnos tics for checking the fit of the imputation model based on comparing i mputed data to nonimputed data. We illustrate with the example that mo tivated this project: a study of pre-election public opinion polls in which not all the questions of interest are asked in all the surveys, so that it is infeasible to impute within each survey separately.