Abstract

The classical discriminate functions have been taught by many researchers, and they found that it is less robustness when some assumptions related with discriminate functions are not achieved specially with the outliers existence, thus the researchers for the last ten years concentrated on robust estimators which was difficult to compute before. But with the development in manufacturing computes from both hardware and software sides, complicated robust estimators became computable and it gives us a new way of dealing with the data compared to the classical estimators . So, the idea of the research is to use robust estimators which are resistant to the outlier influence like robust H estimator, robust S estimator and robust MCD estimator ,and also robust misclassification probability with showing outlier influence on the percentage of misclassification when using classical methods, some of the research aims are to compare estimators to find the best estimator which can give minimum probability of misclassification especially with the variety of contamination percentage and different sample sizes and the data contaminated.