Lundstrom and Sarndal (1999) use calibration to reduce both the sampling error and the nonresponse bias in a survey. The calibration procedure generates final weights which are as close as possible to the design weights, while respecting known auxiliary population totals or unbiased estimates of these totals. A single set of variables is employed to this end and a single calibration step. In this paper, we extend this approach to allow for more flexible implicit modeling of both the response mechanism and the relationship of the response variable with the auxiliary ones. By using Penalized splines the simplicity of the approach in Lundstrom and Sarndal (1999) and the linearity of the estimator is preserved. The technique proposed here can be employed for nonresponse adjustments in a Census and it can be adapted to handle coverage errors, either from frame undercoverage or unit duplication.
Semiparametric Calibration for treatment of Non Response
MONTANARI, Giorgio Eduardo;RANALLI, Maria Giovanna
2008
Abstract
Lundstrom and Sarndal (1999) use calibration to reduce both the sampling error and the nonresponse bias in a survey. The calibration procedure generates final weights which are as close as possible to the design weights, while respecting known auxiliary population totals or unbiased estimates of these totals. A single set of variables is employed to this end and a single calibration step. In this paper, we extend this approach to allow for more flexible implicit modeling of both the response mechanism and the relationship of the response variable with the auxiliary ones. By using Penalized splines the simplicity of the approach in Lundstrom and Sarndal (1999) and the linearity of the estimator is preserved. The technique proposed here can be employed for nonresponse adjustments in a Census and it can be adapted to handle coverage errors, either from frame undercoverage or unit duplication.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.