Thank you Gabor Grothendieck for your message. I would surely like to say, that if someone wants to assume AR(1) residuals, running the regression y ~ x, could run gls(y~x, correlation = corAR1(0, ~1)) Constantine Tsardounis http://www.costis.name