Software for Sample Survey Data, Misuse of Standard Packages

Abstract

Abstract Many health science researchers conduct primary or secondary analyses with data from complex sample surveys but do not realize that specialized statistical software needs to be used. This article compares analyses of survey data conducted with survey software and with standard statistical software. The numerical example illustrates that the inappropriate use of standard statistical software for sample survey data can result in: (a) biased point estimates of population parameters, (b) underestimation of standard errors, (c) confidence intervals that are too wide, and (d) misleading tests of statistical significance. Weighted analyses with standard statistical software, with regular or normed weights, can yield correct point estimates of population parameters but still yield incorrect results for estimated standard errors and statistical tests. Data analysts are advised to use appropriate software for the analysis of survey data.