Consequences of analysing complex survey data using inappropriate analysis and software computing packages

In the analysis of complex survey data such as stratified multi-stage cluster samples, ignoring the design effects such as clustering and stratification usually will lead to erroneous conclusions. In this paper, we will demonstrate the consequences in the estimation of means and proportions by two examples from a stratified two-stage cluster sample. A brief review of methodology will be presented, and some suggestions on computational issues will be provided.