s-news
[Top] [All Lists]

predict interaction in logistic regression

To: <s-news@lists.biostat.wustl.edu>
Subject: predict interaction in logistic regression
From: "GIORGI Roch" <roch.giorgi@mail.ap-hm.fr>
Date: Thu, 29 Jun 2006 18:34:53 +0200
Thread-index: AcabmfOi3W/oYgP6THe/3uTYSF6VAw==
Thread-topic: predict interaction in logistic regression
Dear all,

I am fitting a logistic model : 

glm(y ~ agecl*period + size + gradecl, ...)

where:
agecl  =0 for "aged 50-59", =1 for "aged 60-69", =2 for "aged 70 +"
period  =0 for "1990-92", =1 for "1993-95", =3 for "1996-98"
size  is in a continuous form (from 1 to ...)
gradecl =0 for "Grade1", =1 for "Grade2", =2 for "Grade3"

Question :
The interaction terme is significant. I would like to get estimate for all the 
effects, mainly concerning agecl*period and then to plot the "agecl" effect 
according to the "period". 
I think that I have to use the predict.glm function and have to create a 
newdata containing all the predictors. Given that my focus is on agecl*period 
what values I have to put in these newdata for size and gradecl?

Am I wrong?

Thanks for any comments and help for a computational solution.

Yours

Roch



_____________________________________________________
  Roch GIORGI, MD, PhD
  Service de Santé Publique et d'Information Médicale
  Hôpital de la Timone
  264, rue St Pierre 
  13385 Marseille cedex 5 
  Tel: +33 (0)491 384 949
  Fax: +33 (0)491 385 749 
  email: roch.giorgi@ap-hm.fr 
  Web: http://cybertim.timone.univ-mrs.fr/


<Prev in Thread] Current Thread [Next in Thread>
  • predict interaction in logistic regression, GIORGI Roch <=