I am currently using the relaimpo package to estimate the relative
importance of regressors (N= 4000):
> m1 <- lm(y ~ x1+x2+x3+x4+x5+, data=data)
> calc.relimp(m1, rela=TRUE)
> m2=boot.relimp(m1, boot = 500, rela=TRUE, type="lmg")
> booteval.relimp(m2)
> plot(booteval.relimp(m2))
In a new dataset with 3 measurement points (0,6,12 weeks), I want to
perform a similar analysis, and want overall relative importance estimates
over all 3 time points.
A standard mixed effects model would be adding time as fixed and subject as
random effect:
> m1 <- lmer(y ~ x1+x2+x3+x4+x5+time+(1| subject), data=data)
> m1.p = pvals.fnc(m1)
Unfortunately, that does not allow me to estimate relative importance
because relaimpo cannot handle LME4 output.
> Error in calc.relimp.default.intern(object = <S4 object of class
> "mer">, : If x is NULL, then object must be a data frame or a
> matrix.
Would it be possible to transform the LME4 output in a way that relaimpo
could read it? Do you have other ideas as to how to tackle this, e.g. other
ways of tackling the concept of relative importance / strength of the
regressors?
Thank you
EF
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
