[R] Stepwise regression

[David Winsemius]

On Dec 10, 2008, at 8:03 AM, Frank E Harrell Jr wrote:

> Maithili Shiva wrote:
>> Hi,
>> I have the response variable 'Y' and four predictors say X1, X2, X3  
>> and X4. Assuming all the assmptions like Y follows normal  
>> distribution etc. hold and I want to run linear multiple  
>> regression. How do I run the stepwise regression (forward as well  
>> as the backward regression).
>>> From other software (i.e. minitab), I know only X1 and X2 are  
>>> significant so my regression equation should be Y = constant +  
>>> (b1) X1 + (b2) X2.
>
> Knowing that only X1 and X2 are "insignificant" does not mean you  
> know that b3=b4=0, so why are you deleting X3 and X4 from the model?

I think Frank meant to say "knowing that only X1 and X2 are  
"significant" does not mean you know that b3=b4=0'. His point I am  
guessing, is that when there are valid reasons from prior work  
regarding prediction of Y to think that X3 and X4 ought to be in the  
model, that it is not proper to drop them just because your data does  
not happen show them to be "significant".

-- 
David Winsemius


>
> Frank
>
>> I am not sure whether I am using the correct R commands, which are  
>> as follows -
>> library (wle)
>> ONS <- mle.stepwise(Y ~ X1+X2+X3+X4, data=ONS)
>> summary(ONS)
>> The output looks something like this
>> OUTPUT
>> Forward  selection procedure
>> F.in:  4 Last  3  iterations:
>>     (Intercept) X1 X2 X3 X4      [1,]           1  0  0  0  1 15.57
>> [2,]           1  1  0  0  1 14.80
>> [3,]           1  1  1  0  1 61.56
>> I am not able to interpret this outcome as well.
>> Please guide.
>> With regards
>> Maithili
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>
>
> -- 
> Frank E Harrell Jr   Professor and Chair           School of Medicine
>                     Department of Biostatistics   Vanderbilt  
> University
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.