Mailing this twice ain't going to help you. Reading a course on statistics
might.

The test you want to do is answering following hypothesis : The mean
predicted value of a specific model differs when different datasets are used
to fit it. Seems likely to me if the datasets are not almost identical. Why
testing?

About that Z-test : that should be used in your field of research to test 2
proportions that are not too close to 0 or 1 and that originate from a
binomial distribution with large enough n. Suggesting to use it for
comparing a number of series of around 20 logit-transformed predicted
probabilities is plain shocking.

In case you are interested in the difference of the intercept for these
specific trials, add trial as a fixed effect to your model and do the
appropriate testing. You want to know whether the relation between state and
days differs in slope, you add an interaction term and again use the
appropriate testing. To know what is the appropriate testing, see line 1.

Cheers
Joris

On Thu, Jun 3, 2010 at 10:31 AM, Sacha Viquerat
<sacha.v...@googlemail.com>wrote:

> dear list!
> i have run several glm analysises to estimate a mean rate of dung decay for
> independent trials. i would like to compare these results statistically but
> can't find any solution. the glm calls are:
>
> dung.glm1<-glm(STATE~DAYS, data=o_cov, family="binomial(link="logit"))
>
> dung.glm2<-glm(STATE~DAYS, data=o_cov_T12, family="binomial(link="logit"))
>
> as all the trials have different sample sizes (around 20 each),
>
> anova(dung.glm1, dung.glm2)
>
> is not applicable. has anyone an idea?
> thanks in advance!
>
> ps: my advisor urges me to use the z-test (the common test statistic in my
> field of research), but i reject that due to the small sample size.
>
> ______________________________________________
> 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.
>



-- 
Joris Meys
Statistical Consultant

Ghent University
Faculty of Bioscience Engineering
Department of Applied mathematics, biometrics and process control

Coupure Links 653
B-9000 Gent

tel : +32 9 264 59 87
joris.m...@ugent.be
-------------------------------
Disclaimer : http://helpdesk.ugent.be/e-maildisclaimer.php

        [[alternative HTML version deleted]]

______________________________________________
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.

Reply via email to