[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: dependent variables= mortality rates in GEE and sample size????

From   Maarten buis <>
Subject   Re: st: dependent variables= mortality rates in GEE and sample size????
Date   Thu, 29 Oct 2009 23:21:50 +0000 (GMT)

--- On Thu, 29/10/09, Alison McCarthy wrote:
> I have panel data of 38 countries over the period 1980 to
> 2005. I have calculated age standardized mortality rates in
> each of these countries. Can I use these as the outcome
> variable of internet in my GEE model or is there an issue
> because they are rates? 
> Also is a sample size of 38 countries (obs 748) too small?

I don't think that the sample size is an issue. Actually, it 
is not ideal, but there isn't much you can do about it, as 
you probably got already all countries for which this data
is easily available.

Anyhow the real problem is how you want to interpret the 
results: has internet access an influence on an individuals 
probability of dying (soon) or on the mortality rate in a 
country. These effects can be very different and easily
have opposite signs, so the one says very little about the 
other. Consider for example voting republican or democrat 
in the US: The rich states vote democrat (east coast and
west coast) while the poor states vote republican. However,
rich people are more likely to vote republican and the
poor more likely to vote democrat: so the effect of wealth
on the state level is exactly the opposite of the effect 
of wealth on the individual level. As a consequence, your 
analysis will tell you something about differences
between countries, but it will tell you virtually nothing
about differences between individuals. This phenomenon is 
sometimes known as the ecological fallacy. If you are 
interested in the differences between states then this is 
not a big deal, but if you want to know what is happening
to individuals ...

Hope this helps,

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

Send instant messages to your online friends 

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index