Correcting for Nonresponse in Transition Matrices Calculated from LongitudinalData

摘要

Longitudinal data suffers from the same statistical problems as cross-sectionaldata. True estimates of means and sums requires adjustments for sampling rates or, in the case of census data, nonresponse rates. Whereas in cross-sectional data the standard method of adjustment is the attachment of weights to individual observations, the method does not work in the case of transition matrices calculated from longitudinal data. The reason for this is the unavoidable misclassification of farms as exiters, entrants, and continuers that arises from nonresponse. The method presented here offers a simple algorithm based on four assumptions for making the necessary correction. The algorithm is easily implemented with standard spreadsheet software.