Regression estimation under sampling with one unit per stratum

Abstract

Regression estimation in finite population sampling drawing one unit per stratum is considered in the paper. It is proved that under mild conditions,the generalized regression estimator proposed by Särndal (1982) converges in probability to the population mean, with a rate of convergence of order . Using this result, a variance estimator of the variance of the generalized regression estimator is proposed. Numerical study shows that this variance estimator works better than two alternatives proposedby Särndal.

Keywords:

• Stratified sampling
• Variance estimate
• Linear model
• Covergence in probability