A new method for estimating subgroup means under misclassification

This paper considers the estimation of the subgroup means of a response variable when the individuals in a random sample are allocated to the subgroups or classes by a fallible device. A double sampling scheme under which we determine also the true classes of the individuals in a subsample is utilized. A simple distributional-free estimator is proposed and the associated large-sample properties are also derived. The proposed estimator is also demonstrated by simulation studies to be more efficient than Selén's estimator (1986). Copyright © 1988 Oxford University Press on behalf of Biometrika Trust.