Standard error for an odds ratio (created 2009-08-12)

This page is moving to a new website.

I submitted an article to a journal that included some odds ratios and their confidence intervals. The journal editor said that their policy was to report standard errors and not confidence intervals. How do I do this for an odds ratio?

The odds ratio is by its nature asymmetric, and reporting a standard error that would be expected to be applied in a symmetric fashion is likely to lead to serious problems. In some situations, it might produce a lower limit for the odds ratio that is negative.

You can compute a standard error after using a log transformation. I show the formula for this at my old website:

But the journal that does not like confidence intervals probably would not like log transformations either.

It would also be difficult to adapt the standard error to the situation where the inverse of the odds ratio was preferred. The Cochrane Collaboration, for example, always re-expresses its odds ratios in such a way that a ratio less than one favors the new therapy.

So I'd complain about this journal's policy. I do understand that a table with standard errors is cleaner and simpler than a table with confidence intervals, but the price you pay for this simplicity is too great to tolerate.