A SPLIT QUESTIONNAIRE SURVEY DESIGN

This article develops a survey design where the questionnaire is split into components and individuals are administered the varying subsets of the components. A multiple imputation method for analyzing data fro m this design is developed, in which the imputations are created by ra ndom draws from the posterior predictive distribution of the missing p arts, given the observed parts by using Gibbs sampling under a general location scale model. Results from two simulation studies that invest igate the properties of the inferences using this design are reported. In the first study several random split questionnaire designs are imp osed on the complete data from an existing survey collected using a lo ng questionnaire, and the corresponding data elements are extracted to form split data sets. Inferences obtained using the complete data and the split data are then compared. This comparison suggests that littl e is lost, at least in the example considered, by administering only p arts of the questionnaire to each sampled individual. The second simul ation study reports on the investigation of the efficiency of the spli t questionnaire design and the robustness of the estimates to the dist ributional assumptions used to create imputations. In this study sever al complete and split data sets were generated under a variety of dist ributional assumptions, and the imputations for the split data sets we re created assuming the normality of the distributions. The sampling p roperties of the point and interval estimates of the regression coeffi cient in a particular logistic regression model using both the complet e and split data sets were compared. This comparison suggests that the loss in efficiency of the split questionnaire design decreases as the correlation among the variables that are within different parts incre ases. The proposed multiple imputation method seems to be sensitive to the skewness and relatively insensitive to the kurtosis, contrary to the assumed normality of the distribution for the observables.