A ridge-shrinkage method for range-restricted weight calibration in survey sampling / J. N. K. Rao and A. C. Singh. : CS11-619/97-6E-PDF

"The generalized regression (GR) method is often used to adjust sampling weights to satisfy benchmark constraints (BC), but some of the calibrated weights may not satisfy range restrictions (RR). BC are needed in the interest of efficiency due to correlated auxiliary information and also to make estimates internally consistent with published population totals, while RR are needed to avoid extreme weights which may render domain estimates inefficient. To address this problem, several iterative methods that attempt to meet both RR and BC have been proposed in the literature"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.841397&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs