Impact of correlated misclassifications on polytomous survey data

Abstract

The effect of misclassifications on survey data in the estimation of population proportions in mutually exclusive categories of a polytomous variable is studied under a general misclassification model. The misclassification process is assumed to result in correlated errors. The variance of the estimator for given category, and the covariance of estimators of two different categories are derived. It is shown that under misclassification with no overall bias, the variance of the proportion estimate can be less under misclassifications.

Keywords: Misclassifications, polytomous variable, correlated misclassification errors.


> East African Journal of Statistics Vol. 1 (2) 2006: pp. 130-142