Robust methods of estimation of correlation coefficients (Q580298)

The estimation of the correlation coefficient of a bivariate normal distribution is considered for the case that the sample is ``contaminated'' with data that have a different distribution. A number of conventional robust algorithms are studied, and some new algorithms are proposed. Their properties are examined on finite samples and in the asymptotic case. These algorithms can be used in robust estimation of the covariance matrices and the correlation functions of stochastic processes.