Robust estimation of a correlation coefficient for $\varepsilon$-contaminated bivariate normal distributions

Robust estimations of a correlation coefficient, based on: (i) direct robust analogues of a sample correlation coefficient, (ii) nonparametric estimations of correlation, (iii) robust regression, (iv) robust estimation of the major constituents of a variance, (v) stable estimation of parameters, and (vi) preliminary removal of outliers from data with the following application of a sample correlation coefficient to the residuary observations, are studied. Their application to contaminated normal models on small and large samples is studied, the best robust estimations from the suggested are pointed out.