[R] Strange R squared, possible error

Ben Bolker bbolker at gmail.com
Thu Mar 17 03:11:41 CET 2011


On 11-03-16 07:55 PM, Peter Ehlers wrote:
> On 2011-03-16 15:02, Ben Bolker wrote:
>>   <JLucke<at>  ria.buffalo.edu>  writes:
>>
>>>
>>> lm(y~x+0) yields the regression on x without the constant, i.e., y=bx+e,
>>> not y = a +e
>>>
>>> derek<jan.kacaba<at>  gmail.com>
>>> Sent by: r-help-bounces<at>  r-project.org
>>> 03/16/2011 03:49 PM
>>>
>>
>>    Would someone like to (please!) write this up and submit it to
>> Kurt Hornik for inclusion in the R FAQ?
> 
> Not to belittle Joseph's succinct comment, but I think
> that it's more precise to say that the model through
> the origin is regression with an intercept fixed at zero
> rather than "without the constant". This seems to me
> to be a crucial distinction here and Thomas' post made
> that point very eloquently with:
> 
>   "If you don't know a priori that E[Y]=0 when x=0
>    you shouldn't be fitting a line through the origin."
> 
> Peter Ehlers

  I didn't mean this particular comment: I meant as long a discussion of
the topic as seems appropriate for the FAQ.



More information about the R-help mailing list