[R] How to exclude insignificant intercepts using "step" function

Chris Friedl cfriedalek at gmail.com
Wed Jun 24 01:52:41 CEST 2009




Dieter Menne wrote:
> 
> 
> 
> David Winsemius wrote:
>> 
>> 
>> On Jun 23, 2009, at 3:08 AM, Chris Friedl wrote:
>> The point is that in very few applications can one legitimately  
>> "exclude" an intercept. In this situation (stepwise regression) I am  
>> able to think of a way to make the intercept just another covariate,  
>> but I see theoretic objects with that approach. Of course there are  
>> problems with stepwise regression as well.
>> 
> 
> I agree with David's warning. You should have a-priori knowledge that the
> intercept is zero, not one derived from the significance. For example,
> when your have hill-races (from MASS), where you know that for 0 meter
> (sorry, inches) you need 0 seconds; or from growth curves of artificial
> bone material, where  by definition at t=0 you start with zero bone.
> 
> Dieter
> 
> 
> 

Theoretical considerations would suggest that the intercept is zero. My
intended application is one in mechanical/chemical engineering not so far
removed from zero strain equates to zero stress. Thanks for the warning.







-- 
View this message in context: http://www.nabble.com/How-to-exclude-insignificant-intercepts-using-%22step%22-function-tp24158818p24176497.html
Sent from the R help mailing list archive at Nabble.com.




More information about the R-help mailing list