R help - Jul 2000 - glm and capture-recapture

Hello,

I am almost new in R, so perhaps my question will be silly.

I try to use R for analyzing capture-recapture data in epidemiology. A cancer
registry has different sources of patients. We know in each list, patients
already known in all other list. The aim is to use capture-recapture models for
estimating the number of patients unknow of all the sources.

Because no order in sources, one cannot use classical models for closed animal
populations (models M(bht) of Otis ) so I use hierarchical loglinear model,
moreover for controlling interactions (dependence) between sources. For me it
sound a little strange when I read in papers that authors use 'glm' and
'predict' for this problem even if I know that loglinear are in
generalized linear model.

Why don't use 'loglin'? What is the code for this problem using
'glm' + 'predict'.

I thank you in advance for your help,

==================================Dr SAULEAU Erik-A.
DIM
----------------------------------------------------------
Etablissement Public de Sant? Alsace Nord
141, Ave de Strasbourg
67170 Brumath
Tel : 03-88-64-61-81
E-Mail: dim.brumath at wanadoo.fr
----------------------------------------------------------
Centre Hospitalier d'Erstein
13, Route de Krafft
BP F
67151 Erstein Cedex
E-Mail: dim at ch-erstein.fr
==================================-------------- next part --------------
An HTML attachment was scrubbed...
URL:
https://stat.ethz.ch/pipermail/r-help/attachments/20000725/1e89a309/attachment.html

On Tue, 25 Jul 2000, dim.brumath wrote:
> Hello,
> 
> I am almost new in R, so perhaps my question will be silly.
> 
> I try to use R for analyzing capture-recapture data in epidemiology. A
cancer registry has different sources of patients. We know in each list,
patients already known in all other list. The aim is to use capture-recapture
models for estimating the number of patients unknow of all the sources.
> 
> Because no order in sources, one cannot use classical models for closed
animal populations (models M(bht) of Otis ) so I use hierarchical loglinear
model, moreover for controlling interactions (dependence) between sources. For
me it sound a little strange when I read in papers that authors use
'glm' and 'predict' for this problem even if I know that
loglinear are in generalized linear model.
> 
> Why don't use 'loglin'? 
The user interface to loglin is nigh-impenetrable. `loglm' in package MASS
is a friendlier interface.  But glm is better known and usually
efficient enough.
> What is the code for this problem using 'glm' + 'predict'.
As far as I recall there are several variations on this problem, so 
cannot give you a single answer.
> 
> I thank you in advance for your help,
> 
> ==================================> Dr SAULEAU Erik-A.
> DIM
> ----------------------------------------------------------
> Etablissement Public de Santé Alsace Nord
> 141, Ave de Strasbourg
> 67170 Brumath
> Tel : 03-88-64-61-81
> E-Mail: dim.brumath at wanadoo.fr
> ----------------------------------------------------------
> Centre Hospitalier d'Erstein
> 13, Route de Krafft
> BP F
> 67151 Erstein Cedex
> E-Mail: dim at ch-erstein.fr
> ==================================> 
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._