Re: intercept in multiple linear regression
- From: Bruce Weaver <bweaver@xxxxxxxxxxxx>
- Date: Thu, 14 Jul 2005 09:08:55 -0400
Greg Heath wrote:
thierry wrote:
Hello, is there any difference between regression equation with or without intercept ? I ask this question because I have a negative intercept in my equation (Y = -2000 + 544*var1 + 1166*var2 - 487*var3), in my case intercept doesn't mean anything. If I do an equation without intercept (Y = b1*var1 + b2*var2 + b3*var3), is it correct ?
It is usually recommended that Y and the variables be standardized to have zero mean and unit variance before obtaining the regression coefficients. There will be no intercept if this is done.
Greg, I don't recall seeing standardized linear regression described as the recommended or preferred approach. What are your sources for that?
Thierry, the usual recommendation is to NOT constrain the intecept to be 0 unless there is absolutely no doubt that it has to be 0. (One of the problems is that it messes up interpretation of R^2.) You can find several previous discussions about that in the Google archives.
-- Bruce Weaver bweaver@xxxxxxxxxxxx www.angelfire.com/wv/bwhomedir .
- Follow-Ups:
- Re: intercept in multiple linear regression
- From: Greg Heath
- Re: intercept in multiple linear regression
- References:
- intercept in multiple linear regression
- From: thierry
- Re: intercept in multiple linear regression
- From: Greg Heath
- intercept in multiple linear regression
- Prev by Date: Using null hypothesis testing results in "Bayesian" analysis?
- Next by Date: calculating inflection points of a logit conditional effect plot
- Previous by thread: Re: intercept in multiple linear regression
- Next by thread: Re: intercept in multiple linear regression
- Index(es):
Relevant Pages
|
Loading
