Pseudo-Populations: A Basic Concept in Statistical Surveys

The topic of this book is particularly relevant for statistical surveys, users’ understanding of estimation processes can be substantially improved by generating an artificial pseudo-population for the original population from the sample drawn. Methods such as data imputation and weighting adjustment to compensate for nonresponse can also be explained this way. With or without nonresponse, the actual estimation of the parameter under study is done by calculating this parameter in the pseudo-population. The accuracy of the estimate under different sampling schemes can then be explained by the quality of the pseudo-population generated with regard to the parameter of interest. Another context in which this basic tool from statistical survey analysis offers guidance is the finite population bootstrap. Furthermore, alternative randomized response questioning designs, intended to reduce nonresponse and untruthful answering by increasing the respondents’ privacy protection, can be applied to general probability sampling. Lastly, generating pseudo-populations is very important in the field of statistical disclosure control used by official statisticians.