[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: standard error of the odds ratio
On Monday morning Kate asked about the standard error of the odds ratio:
The epitab command -cc- reports, by default, an exact confidence
interval for the odds ratio, so there is in fact no standard error when
the interval is computed this way.
I need to calculate the standard error of the odds ratio and use the "cc"
command which reports the 95% confidence interval but does not report the
standard error itself. I would appreciate it very much if anyone could let
me know if there is a way or an option to specify that the standard error of
the odds ratio be reported in the output as well.
One thing Kate might want to do is to use the Woolf confidence interval.
One can obtain the upper and lower bounds of the interval using the
option -woolf- on -tabodds-. A standard error is reported if we use the
command -logistic- to obtain the Woolf confidence interval. Here is an
example that shows what I mean by all of this:
cc case exposed
cc case exposed, wo
logistic case exposed
Now, you may notice that the interval is not symmetric so one doesn't
just add and subtract 1.96*se to get the confidence interval. It is in
confidence interval of the log of the odds ratio that is symmteric.
Here is how the above interval is calculated (I'm using the option
-coef- on -logistic- to show the actual coefficients of the model
as opposed to the odds ratio here):
logistic case exposed, coef
display "or = " exp(_b[exposed])
display "lb = " exp(_b[exposed]-invnorm(0.975)*_se[exposed])
display "ub = " exp(_b[exposed]+invnorm(0.975)*_se[exposed])
So what is the standard error that -logistic- reports then? The
standard error of the odds ratio is
display "se(OR) = " exp(_b[exposed])*_se[exposed]
(see [R] logistic, page 67 in version 9 manuals).
* For searches and help try: