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

Re: st: Anova

From   David Airey <>
Subject   Re: st: Anova
Date   Mon, 1 Dec 2008 19:43:56 -0600

If a variance component is negative and a small number, just assume it is not different from zero. But earlier did you not say all your variables were significant? So there may be some problem.


On Dec 1, 2008, at 4:42 AM, aapdm wrote:

Dear John,

Thanks for this, the gfields is indeed very useful! I am now using it and it gives me pretty nice results. However I am surprised that the contribution of one of my explanatory variables is negative (although small). Any idea why this can be the case?

Many thanks, Alice.

--- On Fri, 28/11/08, John Antonakis <> wrote:

From: John Antonakis <>
Subject: Re: st: Anova
Date: Friday, 28 November, 2008, 4:32 PM
Simply estimate:

xi: reg y  x1 x2 x3 x4 x4 x6 i.dumm, beta.

If your variables are either binary or continuous variables
then you can examine the beta coefficient (i.e.,
standardized) for relative impact.

For other ways to look at coefficients download

Also "gfields" is nice for breaking down

If you have categorical data with more than two categories
use the "test" command to test whether the dummies
of that variable dummies are jointly different from zero:
e.g., test I_dumm1 I_dumm2

I guess you could convert that F or chi-square that it
reports to an effect size.

John Antonakis
University of Lausanne
*   For searches and help try:

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index