Online Program

An Introduction to Calibration Weighting for Establishment Surveys

Keywords: Survey weight, Ratio, Poststratification, Raking, Generalized exponential form, Variance.

Calibration weighting is a general technique for adjusting probability-sampling weights to increase the precision of estimates, account for unit nonresponse or frame errors, or force internal estimates to be consistent with external measures. Special cases of calibration weighting include poststratification, weighting-class adjustments for nonresponse, raking (iterative proportional fitting), and separate ratio estimation. Although the literature has primarily focused on linear calibration techniques having close ties to the general regression or GREG estimator, attractive nonlinear calibration techniques have also been developed. The goal here is to provide a general introduction to calibration weighting for establishment surveys using as a motivating example a survey of hospital emergency departments.