The Effect of Complex Sampling on Statistical Procedures in Social Science Research

This talk serves as a general methodological overview of the problems arising from the complex design nature of large social science surveys. We shall begin with an overview of the important features of the linear regression models and assumptions underlying it. Then we shall talk about the main features of the complex surveys such as stratification, clustering, and unequal probability of selection. The meaning and usefulness of these features will be overviewed, their relation to the linear regression assumptions highlighted, and corrections for those features in regression analysis provided. Two examples based on General Social Survey are going to be presented that would demonstrate the magnitudes of the differences between estimators that do or do not take the survey design into account.