DSpace Repository

Calibration and integrated weighting in sample surveys

Show simple item record

dc.contributor.advisor Galpin JS en
dc.contributor.author Neethling A en
dc.date.accessioned 2016-09-22T11:14:10Z
dc.date.available 2016-09-22T11:14:10Z
dc.date.created 1999 en
dc.date.submitted 2004 en
dc.identifier.uri http://hdl.handle.net/20.500.11892/106337
dc.description.abstract The main aim of this thesis is the evaluation of different estimation techniques aimed at creating a single set of weights for persons and households in a household sample survey, such that the resulting estimates are compatible with external population information, and have as small a variance and bias as possible. The emphasis is on the extension of calibration techniques and cosmetic estimators to the integrated weighting technique and the implication as to bias and variance. <br><br> In view of this, theoretical aspects are presented, after which the estimators are applied to a data set and tested using a simulation study. The influence of different sets of auxiliary information - only person level information or both person and household level information simultaneously, is considered in the estimation process of person level and household level study variables. Simple random sample design weights as well as unequal design weights are separately applied. <br><br> The most important findings include the following: (1) The calibration technique can be extended to form so-called calibration-integrated weighting estimators, which reconcile person and household weights to a single set of weights; (2) The cosmetic estimator, in both the design-based and model-based approaches, can also be adapted to yield new integrated-cosmetic estimators; (3) The integrated weighting techniques, in comparison to calibration techniques, in general yield more acceptable variance estimators; (4) Integrated weighting techniques based on both person and household auxiliary information give the best estimators and overcome the problem of inconsistency between person-based and household-based estimators; (5) Only the integrated weighting technique, based on both person and household auxiliary information simultaneously, creates a single set of weights which can be applied to the estimation of person level as well as household level variables, without any rescaling; and (6) The same results are obtained for the application of integrated weighting at the person level data set or household level data set. <br><br> In the final analysis, the study indicates that integrated weighting techniques are important and most useful techniques for the scientific estimation of variables in household surveys, or for any study compiled by different level variables. en
dc.language English en
dc.title Calibration and integrated weighting in sample surveys en
dc.type Doctoral degree en
dc.description.degree PhD (Science) en


Files in this item

Files Size Format View

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record