Re: st: Heavy representation in a sample

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Mon, 21/3/11, Claude Francoeur wrote:
> The total number of observations is 1,103 divided this way
> :  Anglo-Saxon : 730; French 176; German 115 and
> Scandinavian 82.
>
> The sample comprises 21 countries. There are 7 Anglo-Saxon
> and 7 French within the sample. US firms dominate the
> Anglo-Saxon group with 466 firms. This is due to the
> availability of data from public databases.
>
> Since the standard error will be larger, I gather that it
> will be harder to get significant coefficients from the
> regression.Would it be acceptable to reduce the number of US
> firms randomly to avoid this?

No, and that would not help either. The problem is the low
number in the other countries not the high number of US firms.

> On the other hand, if we did find significant effects,
> could we be blamed for the fact that our sample is
> unbalanced?

No, in that case you just found a significant effect using a
test with little power. That is unlikely to happen but when
it does, it is a good thing.

> What would you do in our place?

Just run the analysis I wanted to run, but not over-interpret
non-significant results: those mean an absense of evidence and
not an evidence of absense (this is always the case, but it
is doubly important when you have little statisical power).

-- Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------



      

*
*   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/