Amostragem sistemática com passos modificados

Abstract

Systematic sampling is a sampling method widely used in practice for being simple and fast. However, it presents certain limitations such as: (i) impossibility of estimating variance of estimators from a single sample, and (ii) loss of precision if the population has some measure of periodicity. In this study, we propose an alteration in the method in order to ameliorate loss of precision in these cases while maintaining the simplicity of classic systematic sampling. Therefore, we propose as alteration the introduction of an offset to be added or subtracted to the enumeration, altering which are the units sampled. The class of methods proposed were denominated systematic sampling with modified steps. Through Monte Carlo simulations, we analyzed the precision of alternative types of systematic sampling with modified steps for five different populations (one random population and four periodic populations) in relation to simple random sampling and classic systematic sampling. The sample size was chosen in order to generate worst case scenario in the ordered populations and the random population is considered in this study as a usual situation. Our results suggest that the proposed methodologies result in estimators with better precision than the classic method of systematic sampling for the populations with different periodicities considered.