Re: st: logistic regression confidenceintervals

[Steven Samuels <sjsamuels@gmail.com>]

Yes, to test whether the OR in the shorter period (subset) is equal to the OR over the entire interval, compare the OR in the shorter period to the OR in the complement (all data  1970-99 _not_ in the subset).

Create a variable "period" with 1 = the shorter period of interest  2 = complement

Then rerun your logistic model with a main effect of period and an interaction of period & the predictor that is involved in the OR.

The test of interaction is a test that the OR in period 1 equals the OR in period 2.   If yes, then that common OR is the same for the entire period; if no, then the OR in period 1 is different from the OR in the entire period.

Steve


On Jun 9, 2011, at 12:29 PM, Fitzmaurice, A.E. wrote:

Hi

I have two different models in stata

One covers the period 1970-1999 and gives an odds ratio of 1.35 with a confidence interval of (1.10,1.64)

And the second is a subset of the data covering a shorter time period and gives odds ratios of 1.56 with confidence interval(1.17,2.07)

My question is , is there a way in stata to formally test if the change in the odds ratio is significant,

Thanks

Ann



*
*   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/