Analyzing Survey Data with Weights – A Practical Introduction
Even if it is not apparent to many daily users of statistics, analyzing survey data can be one of the most challenging tasks in inferential statistics. Survey methodologists and statisticians have long acknowledged this pointing to the so-called Total Survey Error (TSE). Although there is an awareness of the complex error structure, all too often statistical inference is done with methods that implicitly assume rather simple errors that are straightforward to estimate. This is particularly true concerning the usage of survey weights. Often considered a nuisance, but necessary for unbiased estimates for so-called descriptive statistics, their effect on standard errors is seldomly considered using the appropriate methodology. The lecture will give basic insights into the construction and logic of survey weights and the appropriate methods are when using them for statistical inference.
