[R-sig-ME] algorithm did not converge when running general linear mixed mode (logistic)
bates at stat.wisc.edu
Thu Nov 6 20:15:32 CET 2008
On Thu, Nov 6, 2008 at 10:19 AM, Miao-Hsuan Yen <melissayen at gmail.com> wrote:
> Dear mixed-list
> I am quite new with the lmer program.
> I have a data set with binary response (1, 0 and NA for missing trials).
> First, I'd like to make sure that it is correct to use the following command
> to analyze such data.
> model = lmer(resp ~ A + B + C + B:C + (1|sub) + (1|item), data = xxx, family
> = "binomial")
> Second, for some instances, I got the warning message:
> In glm.fit(fr$X, fr$Y, weights = wts, offset = offset, family = family, :
> algorithm did not converge
That is an indication that your fixed-effects terms may be
overspecified. To obtain starting estimates for the fixed-effects
parameters a model without random effects is fit initially. It is the
glm(resp ~ A + B + C + B:C, data = xxx, family = "binomial")
that is failing so you may want to check the fixed-effects specification.
> and I got extremely small z values and all Pr(>|z|)s = 1.
> Is there anything I can do to solve this problem?
> Thank you.
> [[alternative HTML version deleted]]
> R-sig-mixed-models at r-project.org mailing list
More information about the R-sig-mixed-models