Re: st: Twin data: descriptive statistics with robust standard deviation

[Stas Kolenikov <skolenik@gmail.com>]

Esther Kessler is asking what kind of corrections are needed for
summary statistics based on twin data.

A mean is a mean, and a standard deviation is a standard deviation.
They have clearly defined formulae, and the results are indeed
delivered to you by -summarize-. What needs correction for clustering
are the standard errors of the parameter estimates. For summary
statistics, you would probably do this with -mean , cluster( pair )-.
That might be what "other researchers" (reference?) mean by using
alternative statistics.

Also, the standard deviation (or rather the variance) may not be
particularly meaningful per se for clustered data, and can be
decomposed into within and between pair component, which would be
easiest done by -anova-. Or at least the SD should be accompanied by
the ICC.

On Wed, Jul 13, 2011 at 1:04 PM, Esther Kessler
<esther.kesslerucl@googlemail.com> wrote:
> Dear all,
>
> I am working with twin data. I want to estimate simple descriptive
> statistics such as mean/stand deviation but i would like to use robust
> standard errors.
>
> Researcher report in publications: that they used robust statistics
> taking dependence of twin data into account when calculating the
> mean/SD of their variables.
>
> Normally I would simple use the command summarize. I could not find a
> suitable option with "summarize". I know the options cluster and
> robust with regressions.
>
> Does anybody know how to calculate this? Has a suggestion?

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   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/