Re: st: xtgee for skewed data

[Nick Cox <njcoxstata@gmail.com>]

The discussion started by William Gould at

http://blog.stata.com/2011/08/22/use-poisson-rather-than-regress-tell-a-friend/

seems relevant. You have a massive spike in the distribution. No
transformation will much affect that, as by Murphy's theorem a spike
maps to a spike, and in any case there would be the usual argument
about what to do with zeros. However, (importantly different here)  a
logarithmic link might help a lot.

The key substantive issue is whether all the people with zeros really
belong in the same model as the others.

Nick

On Thu, Aug 25, 2011 at 11:56 AM, Andre Siqueira
<andresiqueira10@hotmail.com> wrote:

> I’d like some opinions about how to build a model of a very skewed data
> (parasitemia) using GEE.
> My data consist of repeated measures of parasitemia of many individuals
> (some of them having only one measurement and others having up to five) I’d
> like to assess the association with a 4-level exposures adjusted for 3
> covariates. Only 5% of the measurements had a value above zero.
> I’ve already discussed with some people who told me not to transform the
> data and analyse it with a normal distribution, so I’ve built the following
> model but I’m still in doubt if it’s the best way to do it:
>
> Xtgee parasitemia i.main_exp i.co_var1 i.co_var2 i.co_var3, i(id)
> vce(robust)
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/