DOI: 10.5176/2251-1938_ORS18.07
Authors: Omer Ozturk
Abstract:
In this paper, we construct a sampling design to improve the information content of a simple random sample (SRS). We rst start with a simple random sample of size n. By condition- ing on the population rank of each one of the n units in SRS, we nd the conditional distribu- tion of the sample rank of each measured obser- vation in SRS among all units in a comparison set of size M, M 1. We then use this condi- tional distribution to post-stratify the sample into d strata, 1 d M. We construct esti- mators for the population mean and total and their variances. For certain choices of d, the estimators of population mean and total out- perform the same estimators constructed from a systematic sample. The paper shows that all estimators are unbiased.
Keywords:Systematic sample; post- strati ed sample; judgment post-strati ed sample; Rao-Blackwell estimator; cluster sam- ple; ranked set sample
Updating...