posted and e-mailed. On Wed, 07 Jan 2004 14:43:09 -0500, Bruce Weaver <[EMAIL PROTECTED]> wrote:
> Rich Ulrich wrote: > > ---- snip ------ > > > > Also, a few decades ago, heightened interest in health > > (tobacco, asbestos, industrial exposures) meant that > > "survivorship curves" grew in popularity, along with several > > related concepts including relative-risks. However, true > > "relative risks" are relatively intractable in math, compared > > to odds-ratios, so the latter have been given a bigger stage. > > And it sometimes happens that Odds ratios are mislabeled > > Relative risks, because the latter name seems more 'natural.' > > > > There may be another reason for someone calling an odds > ratio a risk ratio (or relative risk). I have a chapter by > Hennekens & Buring (1987) that describes the analysis of > case-control studies. The authors give the formula for the > odds ratio, but call it the relative risk. I have always > assumed they did this because the OR from a case-control > study is a pretty good estimate of the RR in the population > from which one has sampled. Didn't Mantel/ Haenszel provide the precedent? Their MH Relative Risk for stratified 2x2 tables was an early use of OR, and I don't think of one earlier. - Yes, you are right. For small ratios, the OR estimates the RR. In addition to what I said before about the RR sounding natural: It might be ungenerous of me, but I've always assumed that M-H did not want to get into arguments about the OR, so they camouflaged what they were doing. And they could figure that any complaints would be muffled, since the OR has all those elements of natural superiority over the RR when it comes to doing the math or comparing the numbers across contexts. Is this right, where I assume that the Relative Risk was the measure that epidemiologists were accustomed to, and that the OR hardly existed for them, before 1959? -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html "Taxes are the price we pay for civilization." . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
