> It's just by chance that this happened: the > correlation you generated > was not strong enough, or was far from an > unstructured correlation and > the algorithm did not converge (and you obtained > this error message). > Generate another sample might solve the problem. >
I think that you help me on the right point. Since nlme() uses EM and Newton-Raphson in sequence, other than that PROC MIXED uses Newton-Raphson only, as I know, I am wondering whether PROC MIXED can circumvent this situation. But the problem is as what I mentioned in the first mail, which is I do not know how to structure the var-cor with unstructured lower triangle and identical diagonal in PROC MIXED. > You might try "for fun" but it does not make much > sense to me. I am now doing a project which needs to generate random samples, and fit them using mixed models with different var-cov structures. If the problem that I met is common, then I need to revise my codes a little bit. Thank you for breaking my problem. Peng ------------------------------ Peng Liu | Division of Statistics | Northern Illinois University | De Kalb, IL 60115, USA | E-mail: [EMAIL PROTECTED] | ------------------------------ ______________________________________________ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
