A Contamination Model and Resistant Estimation Within the Sampling Paradigm
将Copas的二元回归污染模型转化为判别分析中熟悉的抽样范式，通过统一的似然表达式处理两种模型，并讨论患病率及判别函数参数的稳健估计，以乳腺癌数据为例说明。
SUMMARY Copas's model for contamination in binary regression structures is translated into the sampling paradigm approach that is familiar in discriminant analysis. A parallel treatment of the two models is afforded by a common expression of the log-likelihood. Resistant parameter estimation is discussed for prevalence rates and for parameters, including discriminant functions, within component distributions. The methods are illustrated on a breast cancer example.