[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: AW: st: AW: beta coefficients for interaction terms

From   Maarten buis <>
Subject   Re: AW: st: AW: beta coefficients for interaction terms
Date   Mon, 22 Jun 2009 09:33:24 +0000 (GMT)

--- On Sat, 20/6/09, wrote:
> Why do I have to standardize the dependent variable as
> well? If I standardize it, the constant won't be
> significant anymore. Without standardizing the constant
> is highly significant in my case.

This is not surprising, actually we a-priori know that that
constant should be exactly zero (which is the null-hypothesis
that the reported significance level tests).

This is easiest to see in a bivariate regression between y 
and x. In this case there a couple of properties that are

1) If we standardize a variable then the mean will become 0. 

2) If we fit a regression then the regression line will go 
through the point [mean(x), mean(y)]

3) The constant is the y-value on the regression line when
x equals zero

Combining properties 1) and 2) we know that a standardized 
regression line will go through the point (0,0). Combining
this with propery 3) shows that the constant of a 
standardized standardized regression line is by definition
equal to zero. 

To answer your question, I would say that if you want to 
report standardized coefficients you should standardize 
both the dependent and indpendent variables, not because 
there is some magical/statistical Law that proscribes it, 
but because this is what is meant with standardization in 
most disciplines. If you do it differently you will just 
cause confusion with your readers.

-- Maarten

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen



*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index