[R-lang] Re: % of total variance accounted for by each predictor?
Alex Fine
afine@bcs.rochester.edu
Tue Aug 3 09:24:32 PDT 2010
Just given the latency to first response, I'll give you my best guess:
as far as getting an R^2 measure (=percentage of variance explained)
from a multilevel model, I know of this:
http://hlplab.wordpress.com/2009/08/29/nagelkerke-and-coxsnell-pseudo-r2-for-mixed-logit-models/
As far as getting R^2 for an individual predictor X1, assuming the
pseudo R^2 measure you get from the method described in the blog post
behaves the same as a regular R^2 measure, you could simply take the
pseudo R^2 from a model with X1 + X2 and subtract from that the pseudo
R^2 of a model with just predictor X2. That should give you the portion
of R^2 accounted for by just X1 (i.e. the so-called semi-partial of X1).
alex
Xiao He wrote:
> Hi dear R-lang experts and users,
>
> I have a question regarding the % of total variance account for by
> individual predictors.
>
> In some articles I've read that mention regression analysis (mainly in
> the 2nd language acquisition literature), the authors often mention
> the percent of variance is explained or accounted for by each
> independent variable. In Baayen (2008, in chapter 7), he discusses
> that total variance accounted for by predictors of interests can be
> obtained by abstracting variance of random effects from the variance
> obtained in the model fitted with predictors of interests.
>
> I wonder however if there is any way to tease apart the contribution
> made by each individual variance. In other words, what % of total
> variance is accounted for by each predictor?
>
> Thank you in advance for your help!
>
>
>
> Best,
> Xiao
>
>
>
>
More information about the ling-r-lang-L
mailing list