On the issue of calibrating and weighing data from large-scale sample surveys

The paper deals with the problem of calibrating sample population statistics to known population totals. A comparison analysis is made of various popular weighting methods, with a summary of their merits and shortcomings, and appraisal of their technical calculating difficulty. A new calibration method is proposed, which is based on margin adjustment iterations, similar to those used in the IPFP method, but with floating calibration margins in the contingency table matrix rows. Proof is provided for the convergence of the new method, with the resulting weights all being positive values (in contrast to the linear method of CALMAR).