Path: news.mathworks.com!not-for-mail
From: <HIDDEN>
Newsgroups: comp.soft-sys.matlab
Subject: Re: glmfit binomial-probit with natural mortality
Date: Mon, 29 Sep 2008 23:48:02 +0000 (UTC)
Organization: UCV
Lines: 52
Message-ID: <gbrpfi$dlg$1@fred.mathworks.com>
References: <g6vilo$h12$1@fred.mathworks.com> <gbqs61$9ts$1@fred.mathworks.com> <gbr779$n6e$1@fred.mathworks.com>
Reply-To: <HIDDEN>
NNTP-Posting-Host: webapp-03-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1222732082 14000 172.30.248.38 (29 Sep 2008 23:48:02 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Mon, 29 Sep 2008 23:48:02 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 92693
Xref: news.mathworks.com comp.soft-sys.matlab:492818


Hi Tom

Thanks for your reply. There exists a "standard" statistical model for this type of fit. It is called generalized linear models. This kind of statistical models can deal with a great variety of different problems and parameters. Our particular problem has a non-zero parameter (wich you correctly discribe as a baseline probability). In the biostatistics field (dose-response analysis) is called control mortality or mortality in the control group. I know that programs such SPSS or SAS can handle this type of problems directly. I would like to be able to do it in MATLAB, but I don't know how.

Thanks in advance

"Tom Lane" <tlane@mathworks.com> wrote in message <gbr779$n6e$1@fred.mathworks.com>...
> Joe or Bego, I'm not aware of what standard mathematical/statistical 
> approach that is generally used here.  I could guess that it might involve 
> estimating a non-zero baseline probability of response, with the link 
> function describing how the probability increases.  But I really don't know.
> 
> I guess what I'm asking is this.  Are you looking for the right approach, or 
> do you know the approach and you're looking for advice on how to do it in 
> MATLAB?
> 
> -- Tom
> 
> "Villamide Bego" <bvillamide@gmail.com> wrote in message 
> news:gbqs61$9ts$1@fred.mathworks.com...
> > Hi Joe,
> >
> > I have the same problem as you. Have you any solution to this question?
> >
> > Thanks in advance.
> >
> > Bego.
> >
> >
> >
> > "Joe Ercolino" <joe_ercolino@hotmail.com> wrote in message 
> > <g6vilo$h12$1@fred.mathworks.com>...
> >> Hi
> >>
> >> I'm trying to fit a generalized linear model (GLM) with the
> >> glmfit function. My problem deals with the
> >> dose-response(Mortality)of a population of animals exposed
> >> to a toxic agent. The most common models for this type of
> >> regression use the binomial distribution and probit or logit
> >> as link functions with the log(Dose) as predictors. The data
> >> I'm working with shows a non zero response for the controls
> >> (dose equal to zero). I think that the logit or probit
> >> functions can not accommodate for this type type behavior
> >> since they go from zero to one. Is there a way to make
> >> glmfit to take into account the controls correction (natural
> >> mortality) or there is any other feasible way to do it in
> >> MATLAB?
> >>
> >> Thanks in advance.
> > 
>