Keywords: benchmarking, calibration, range restrictions
In survey weight calibration, it is often the case that no single set of weights can meet all population total constraints while simultaneously satisfying prescribed ranges needed to control for extreme weights. We present a framework and accompanying computational methods to address this issue of constraint achievement or selection within a restricted space that will produce revised weights with reasonable properties. We discuss different components of the framework and compare the use of a logistic measure to that of general quadratic programming. We demonstrate these alternative methods for post-stratification for the National Survey on Drug Use and Health. We also discuss strategies for scaling up to even larger data sets. Results suggest that some quadratic programming problems can more efficiently be approximated with a different objective function (non-linear programming) while still incorporating the original equalities and inequalities.