Dear Igors On 9 September 2011 22:00, Igors <igors.lahanc...@gmail.com> wrote: > Thank you for your answer. However it doesn't solve my problem fully. > > The problem is that I have much bigger data set than I sent to you (it was > only a small part : 874 obs.). My full data set is 546718 obs. > > If I try to use censReg on full data set, then it still gives me the same > already mentioned error about Na's in the initial gradient. > > > I have sent you an e-mail with full dataset and the code.
Thanks! > I would really > appreciate if you could check how it works and suggest me how to solve this > problem. Now, I get the same error message. The problem is that the log-likelihood function becomes minus infinity. I guess that this problem occurs, because a likelihood contribution of (at least) one individual observation is so small that it is rounded zero and hence, its logarithm is minus infinity. However, I have to take a deeper look into this issue before I can give a definite answer and hopefully find a solution. > I have noticed that you use "iterlim" argumet to specify maximum number of > iterations. How this argument could affect possibility of obtaining > estimates? Using your small data set, the maximization of the (log) likelihood function in censReg() did not converge within 100 iterations. Therefore, I increased the maximum number of iterations from 100 to 200 -- and then the maximization converged :-( /Arne -- Arne Henningsen http://www.arne-henningsen.name ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
