Skewed survey populations : optimal construction of "take-all" and "sampled" groups with application to the local government finance survey re-design / Gurupdesh S. Pandher. : CS11-613/95-1E-PDF

"In a survey re-engineering context, the most optimal sample design and estimation strategy holds the promise of offering the largest reduction in the sample size (and survey costs). In this paper, a scheme called the "Transfer Algorithm" is proposed to address the problem of finding an optimal demarcation between the "take-all" and "sampled" groups in surveys of skewed populations (eg. business, agricultural populations). The criterion for constructing these groups is based directly on minimizing the design variance of the regression estimator under a flexible range of sample designs (eg. SRS, pps, stratified). The proposed method also explicitly captures the size-induced heteroscedasticity evident in skewed populations in the determination of the optimal demarcation"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.837182&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs