Complex sampling designs and statistical issues in secondary analysis.

Abstract

Conducting secondary analysis using large national survey data sets to answer pressing research questions is gaining acceptance in the nursing science community. There are, however, challenges confronted by researchers who wish to apply secondary analysis to large data sets due to the incorporation of complex sampling designs. This article presents sampling design issues inherent in many large national surveys and explains the rationale for applying sample and variance estimation weights when conducting statistical analyses. In addition, the rationale for using statistical software packages capable of analyzing data derived from complex sampling designs is described. Examples of differences in statistical outcomes with and without weights using Stata and SPSS are provided using data from the Medical Expenditure Panel Survey (MEPS). Based on the example analyses, the implications of the statistical outcome differences for study findings are discussed.