R help - Sep 2010 - modeling variance heterogeneity in lme4

Hi all,

I have major heterogeneity in variances across labs (100-fold). There is no
apparent variance heterogeneity across y-hat. By using lme4 in the following
way, am I accounting for the variance differences in labs?:

lmer(y ~ fixed1 + covariates + (fixed1|labs))

I'm not sure that it is - I think it is only allowing the means (slopes
[conditional means] & intercepts) to differ between labs. If so, how does
one appropriately model such variances differences?

Thank you in advance,

JJ

	[[alternative HTML version deleted]]

I should have been more careful in my description. When I say there is a
100-fold difference in variance across labs, I mean that variance in fixed1
(the predictor) varies 100-fold across labs, not the variance in y is
100-fold difference across labs.

Can lme4 account for such variance differences?

On Fri, Sep 17, 2010 at 6:43 PM, Johan Jackson
<johan.h.jackson@gmail.com>wrote:
> Hi all,
>
> I have major heterogeneity in variances across labs (100-fold). There is no
> apparent variance heterogeneity across y-hat. By using lme4 in the
following
> way, am I accounting for the variance differences in labs?:
>
> lmer(y ~ fixed1 + covariates + (fixed1|labs))
>
> I'm not sure that it is - I think it is only allowing the means (slopes
> [conditional means] & intercepts) to differ between labs. If so, how
does
> one appropriately model such variances differences?
>
> Thank you in advance,
>
> JJ
>
	[[alternative HTML version deleted]]