@MISC{_lettersto, author = {}, title = {Letters to the Editor Letters to the Editor LETTERS TO THE EDITOR RE: “PRESENTING STATISTICAL UNCERTAINTY IN TRENDS AND DOSE-RESPONSE RELATIONS”}, year = {} }

Share

OpenURL

Abstract

The recent commentary by Greenland et al. (1) criticizes the “floating absolute risk ” (FAR) method for presenting uncertainty in estimates of polytomous exposures. However, it then goes on to propose “floating trends, ” which are exactly the same, to within a constant! The criticism that Greenland et al. raise of the FAR confi-dence intervals is that they are not, directly, confidence inter-vals for the relative risk estimates for each category relative to the baseline category. This is obviously true—they are not meant to be. The whole point of the FAR method is to allow computation of confidence limits for relative risks for any pair of categories. The variance of the log relative risk is obtained by adding the “floating variances ” for the two categories, in exactly the same way as one does in, for example, an unpaired t test.