Don’t Weight the Rates in Nonresponse Weights!
Presented by:
Roderick J. Little & Sonya Vartivarian
University of Michigan
Thursday, April 19, 2001
Abstract
A basic survey estimation strategy is to weight responding units inversely proportional to the probability of selection and response. The response weights are usually inverses of the sample-weighted response rates in adjustment cells, that is, the ratio of the sum of the sampling weights of respondents to the sum of the sampling weights for respondents and nonrespondents in a cell. We show by theory and simulations that the use of sample-weighted response rates is either incorrect or unnecessary. The correct weighting approach is to model nonresponse as a function of the adjustment cell and design variables, and to estimate the response weight as the inverse of the estimated response probability from this model. This approach can be implemented by creating adjustment cells that include design variables in the cross classification, if the number of cells thus created is not too large. Otherwise, response propensity weighting can be applied.