Re: st: Conditional logistic regression

[Steve Samuels <sjsamuels@gmail.com>]

I believe this kind of design is called the "single-subject",
"case-time", or "case-crossover" design, the former in the social
sciences, the latter in epidemiology. I've never done one, but a Google
Scholar search should lead you to references. Off-hand, I see no technical
problem with conditional logistic approach per se. However because, in
your study, the change goes in only one direction, the effect of
pollution will be confounded with temporal trends.

A possibly useful reference, which I haven't read:
Maclure, M., and and M. A. Mittleman. 2000. Should We Use a
Case-Crossover Design? Annual Review of Public Health Annu. Rev. Public.
Health. 21, no. 1: 193-221.


Steve
sjsamuels@gmail.com



On Nov 11, 2013, at 7:43 AM, Gareth Davies (Public Health Wales) wrote:

Dear all

We have some data on all the individuals in a population for two periods: a period during which an environmental pollution event occurred and the same period the previous year when no pollution event occurred.

For this population we know whether the individuals had a health care event, such as an emergency hospital admission, during either period.

We also know other characteristics such as age, and pre-existing co morbidity.

We want to establish whether the pollution event had any effect on the health care events.

Would it be correct to use conditional logistic regression (clogit) to analyse these data on the basis that they are akin to 'matched case-control' data?

Many thanks

Gareth

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