R help - Aug 2003 - Strange predicted values ?

Hello

I carried out a logistic regression and found predicted values.
Then I want to see both predictors (var1,var2..) and predicted values
in same matrix. In other words, I need to know each combinations
and predicted values.

I used:
cbind(var1,var2,var3,var4,predict(glm.obj,type="resp"))

I got  a somewhat strange result:

var1  var2  var3  var4  var5 var6      predicted vals
------   -----  ----- -------- -----  -----      -------------------
  6        6       1      1        1      1            4.24e-07
  6        6       1      1        1      1            8.37e-11
  6        6       1      1        1      1            6.8e-07
   .
   .
   .

I had expected to have a same predicted value for
same combinations.

Can anyone explain to me why ?
Or any suggestion ?

kind regards


Ahmet Temiz
TURKEY


 


______________________________________



______________________________________
The views and opinions expressed in this e-mail message are ...{{dropped}}

Those are not predicted values, they are fitted values.  Try predicting on 
the same set of variables as you printed.

Please do try to give a small reproducible example so we can see what you 
actually did.

On Thu, 7 Aug 2003, orkun wrote:
> I carried out a logistic regression and found predicted values.
> Then I want to see both predictors (var1,var2..) and predicted values
> in same matrix. In other words, I need to know each combinations
> and predicted values.
> 
> I used:
> cbind(var1,var2,var3,var4,predict(glm.obj,type="resp"))
> 
> I got  a somewhat strange result:
> 
> var1  var2  var3  var4  var5 var6      predicted vals
> ------   -----  ----- -------- -----  -----      -------------------
>   6        6       1      1        1      1            4.24e-07
>   6        6       1      1        1      1            8.37e-11
>   6        6       1      1        1      1            6.8e-07
>    .
>    .
>    .
> 
> I had expected to have a same predicted value for
> same combinations.
Yes, so it's a fair guess that is not what you computed.  But, we can't 
tell what you did do.


-- 
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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Prof Brian Ripley wrote:
>On Thu, 7 Aug 2003, orkun wrote:
>
>[quoting me without attribution]
>
>  
>
>>>Those are not predicted values, they are fitted values.  Try
predicting on
>>>the same set of variables as you printed.
>>>      
>>>
>
>Precisely!  From ?predict.glm
>
> newdata: optionally, a new data frame from which to make the
>          predictions.  If omitted, the fitted linear predictors are
>          used.
>
>[...]
>
>  
>
>>If  predict(glm.obj,type="resp") does not give predicted vals,
How can I
>>get predicted values ?
>>    
>>
>
>Try reading the help page?  It is quite explicit, and has examples, as do 
>all good books on S/R.
>
>  
>
I tried this:
#I think since an interaction exists in glm.obj, data.frame.obj did not 
not work
#instead I used model.frame obj (it works)
newdata<-model.frame(glm.obj)
pr<-predict.glm(glm.obj,newdata,type="resp")
it works but there was a warning message:
prediction from a rank deficient fit may be misleading in predic.lm (....)

any suggestions ?

thank you again


______________________________________



______________________________________
The views and opinions expressed in this e-mail message are ...{{dropped}}