Hi. I have a data matrix of size 3072x1910. I can compute a standard 1910x1910 covariance matrix for this, and it comes out positive definite. I wanted to compute a robust covariance matrix with cov.nnve (in the covRobust library). It failed with the following error message: Error in while (abs(loglik.new - loglik.old)/(1 + abs(loglik.new)) > convergence) { : missing value where TRUE/FALSE needed In addition: Warning message: value out of range in 'gammafn' So I'm seeking advice / assistance. I have also tried covRob (in the robust library), which has some different methods for robust covariance estimation, but it runs out of memory (I usually have about 1.5Gb available). -- TMK -- 212-460-5430 home 917-656-5351 cell