[The importance of the sample design effect]

Abstract

Introduction: Sample selection through a complex design influences the subsequent statistical analysis.

Material and method: The different means of sample selection may result in bias and greater variance of estimators; simple randomized sampling is the reference design. Diverse examples are provided, illustrating how the various sampling strategies can result in bias and increase variance.

Results: The inclusion of different weighting techniques reduces bias. Evaluation of the effect of design enables measurement of the degree of variance distortion due to the sampling design used and therefore provides a direct evaluation of the alteration in the confidence intervals estimated when the sampling design deviates from simple randomized sampling.

Conclusions: We recommend measurement of the effect of the design on analysis of the data obtained by sampling and inclusion of weighting techniques in statistical analyses.