Message-ID: <XFMail.080321165400.Ted.Harding@manchester.ac.uk>
Date: 2008-03-21T16:54:00Z
From: (Ted Harding)
Subject: warning message in a glm model
In-Reply-To: <BLU113-W4701B6172D74FEABAE0436A0010@phx.gbl>
On 21-Mar-08 15:59:05, Alfonso P?rez wrote:
> Hello, my name is Alfonso. I want to apply a logistic model to my data.
> The variables are, fish length and age length as a continous variable,
> and cohort as a factor. I consider the interaction between age and
> cohort in the model. The dependent variable is the probability of being
> mature. Then the model is as follow:
>
> ogive <- glm(mature~length+(cohort*age), family=binomial,
> data=subset.cod)
>
>
> But when I run it in R I obtain a Warning message:
>
> Warning message:
> In glm.fit(x = X, y = Y, weights = weights, start = start, etastart =
> etastart, :
> fitted probabilities numerically 0 or 1 occurred
>
> I don't understand what's the problem, could you help me? thank you
> in advance.
>
> Alfonso.
There was effectively the same question from Werner Wernersen:
"[R] glm.fit: "fitted probabilities numerically 0 or 1 occurr"
11 March 2008 08:58:55.
See my response to him:
"Re: [R] glm.fit: "fitted probabilities numerically 0 or 1 occurr"
From: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Date: Tue, 11 Mar 2008 10:08:09 -0000 (GMT)
which covers the maun issues.
The underlying reason is that some linear function of your
covariates compeletly separates the "Y=1" cases from the
"Y=0" cases, so the GLM can predict probability 1 for "Y=1",
and probability 0 for "Y=0".
Best wishes,
Ted.
--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 21-Mar-08 Time: 16:53:54
------------------------------ XFMail ------------------------------