ANALYSIS OF CATEGORICAL DATA WITH MISCLASSIFICATION ERRORS by CHUN-

Errors in the collection of data are obstacles to analysis because the underlying misclassification mechanism is usually unknown. Many authors have investigated this problem. In this paper, we shall concentrate on the analysis of contingency tables which are subject to misclassification errors. A general misclassification framework for multi-dimensional contingency tables is proposed. Base on this framework, a family of misclassification models is generated by imposing sets of constraints which make the parameters identifiable. Log-linear models are considered and iterative weighted least squares method is utilized to find the maximum likelihood estimates of the parameters. The estimated expected cell frequencies are then used to test the goodness-of-fit of the model. In order to partially resolve the difficulties involved in inference from a sample of erroneous categorical data, a double sampling approach is considered.