Hi,
very likely your data exhibit quasi-separation which cause (log)Lik to
be monotone and thus ML estimate do not exist. However you can rely on
point estimate and use LRT to test for its significance.
Or Better: have a look to brlr or logistf packages which bypass the
monotone-likelihood problem by using penalized likelihood.
Best,
vito
Taka Matzmoto wrote:> Hello R users
>
> I ran more than 100 logistic regression analyses. Some of the analyses gave
> me this kind warning below.
>
> ###########################################################
> Warning messages:
> 1: algorithm did not converge in: glm.fit(x = X, y = Y, weights = weights,
> start = start, etastart = etastart, ...
> 2: fitted probabilities numerically 0 or 1 occurred in: glm.fit(x = X, y =
> Y, weights = weights, start = start, etastart = etastart, ...
> ###########################################################
>
> For those cases for which I got the warning messages, shouldn't I rely
on
> coefficents ? It looks like I can still extract coefficients from R outputs
>
> Are there any ways to avoid these warning messages ? or are these due to
the
> problems with my data (e.g., perfect separation)
>
> Any help or advice would be appreciated
>
> Thank you
>
> TM
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>
--
===================================Vito M.R. Muggeo
Dip.to Sc Statist e Matem `Vianelli'
Universit?? di Palermo
viale delle Scienze, edificio 13
90128 Palermo - ITALY
tel: 091 6626240
fax: 091 485726/485612