Reamostragem bootstrap em amostragem por conjuntos ordenados e intervalos de confiança não paramétricos para a média

Abstract

Ranked set sampling is an efficient and practice way to obtain more precise estimative when the sample size is small because of the high cost or difficulties to measure the interest variable. Using rough and cheap qualitative or quantitative information, the sample units are ranked before their effective measurement. In 1952, McIntyre introduced the ranked set sample design to estimate the average yields from plots of cropland, using the ranked set sample mean, X . Cesario and Barreto (2003) have shown a parametric version of bootstrap confidence intervals for normal distribution mean. Because of the restriction of small sample size, the distributional assumption may not be reasonable, producing no liable estimates. So the study and proposition of precise interval estimators of the population mean could be relevant and are the main interest of this work. Using resampling methods, we propose in this work an extension of bootstrap resampling for ranked set sampling. A simulation study is conduced to the properties of single random sample bootstrap confidence intervals and the similar using our version for ranked set sampling. The analysis of the simulation study have shown the gain of precision for using the ranked set sampling bootstrap confidence intervals in the population mean.