Statalist


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

Re: st: R: question concerning clustering - second try


From   Johannes Schoder <johannes.schoder@soi.uzh.ch>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: R: question concerning clustering - second try
Date   Fri, 18 Sep 2009 10:51:22 -0400

Thanks a lot for the reference and your feedback Carlo!
So I am not that wrong with clustering my s.e. for county.
Best,
Johannes

Carlo Lazzaro schrieb:
Johannes wrote:

< I assume that I have to cluster my standard errors since the
observations within county i might be correlated within counties?>

I don't think that Johannes'question is trivial.
As far as I understand your research purpose, I would advise you to cluster
standard errors for county, since individuals living in the same county are
likely to share the same environment and healthy or unhealthy lifestyles,
which may have some influence on mortality rates.

For a more detailed discussion on this topic, I would refer you to:
Kirkwood BR, Stern JAC. Essential medical statistics. Second edition.
Malden, Mass: Blackwell Science, 2003: 355-370.
Although focussed on clinical trials, the reported examples can be easily
exported to other research fields and tweaked accordingly.

Kind Regards,
Carlo
-----Messaggio originale-----
Da: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di Johannes Schoder
Inviato: venerdì 18 settembre 2009 15.44
A: statalist@hsphsun2.harvard.edu
Oggetto: st: question concerning clustering - second try

Dear Statalist Users:

I haven't received any feedback so I hope my question is not too stupid:
I am estimating the impact of the number of deaths per county and cause
of death on the age at death per county and cause.
I have information for each county: on the number, the mean age at
death,  and the cause of death.
The model looks like that:

xi: reg  MeanAge_death log_n_death [aweight=n_death]

where MeanAge_death is the average Age at death in county i due to cause c
n_death: Number of death in county i due to cause c

I assume that I have to cluster my standard errors since the
observations within county i might be correlated within counties?
Then my model would simply look like:

xi: reg  age_death_i,c log_n_death [aweight=n_death],  robust
cluster(county)

But I am not sure if I also have to cluster according to different cause
of death?
Thanks a lot for any suggestion.
Johannes


*
*   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/



*
*   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/

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index