[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Anova
David Airey <firstname.lastname@example.org>
Re: st: Anova
Fri, 28 Nov 2008 09:04:28 -0600
No, I meant the difference in R-squared between the model with all variables vs the one with those of interest removed. But are you really interested in percent variation explained or a measure of effect size? It might also be worth interpreting the coefficients in terms of change in Y with a standardized change in X.
Sent from my iPhone
On Nov 28, 2008, at 8:35 AM, aapdm <email@example.com> wrote:
Thanks - so are you suggesting that I should regress Y on each of
the explanatory variables separately and look at the R2 in each case?
--- On Fri, 28/11/08, David Airey <firstname.lastname@example.org> wrote:
From: David Airey <email@example.com>
Subject: Re: st: Anova
Date: Friday, 28 November, 2008, 2:22 PM
This is true for balanced factorial ANOVA, but probably not
in your complicated model.
For a given variable, why not look at adjusted R^2 with
that variable (or group of dummies if categorical) in an out
of a regression model?
On Nov 28, 2008, at 7:08 AM, aapdm wrote:
I am trying to use the anova command but I am not sure
I am doing the right thing.
I have a dependent variable Y which I explain by 10
explanatory variables, half of which are categorical while
the others are continuous.
When I use the anova command and specify which
variables are continuous, then I get a table with the
Partial SS for each of the explanatory variables.
If I sum the Partial SS for all variables then this is
much smaller than the value reported for the Model SS. How
is that the case? What am I missing here?
What I want to is to find to what extent each of the
different explanatory variables explains the variance of the
dependent variable, which should be given by the ratio
between the Partial SS of each variable and the total SS. Am
* For searches and help try: