 |
 |
Hi Barney and the list, I tried a few calculations on the site and noticed that when I put 3 in the FN place instead of 0, I got a 95%CI that was greater than 1, clearly an "impossible" situation for sensitivity. I assume that the CI calculator for this is not geared to recognize the acceptable upper limit of sensitivity as 1. I would propose that the way to get around this is to calculate the lower limit of the 95% CI using the 3/n rule and then use that to calcuate your LR. After all, the upper limit will continue to be 1. Dan **************************************************************************** Dan Mayer, MD Professor of Emergency Medicine Albany Medical College 47 New Scotland Ave. Albany, NY, 12208 Ph; 518-262-6180 FAX; 518-262-5029 E-mail; [log in to unmask] **************************************************************************** >>> Barney Eskin <[log in to unmask]> 04/07 9:25 AM >>> This question is about calculating 95% confidence intervals for LRs. I use the UBC calculator which usually works fine and is found at http://www.healthcare.ubc.ca/calc/bayes.html If you use the "I know the raw data of the 2 x 2 table" however, and one of the cells has a "0" in it, it does not give me 95% confidence intervals, just a point estimate. Now, I know that you can estimate the limit of the value in the "0" cell by using the approach in the paper by Hanley and Lippman-Hand, If nothing goes wrong, is everything all right? Interpreting zero numerators. JAMA 1983;249:1743-5. That paper suggests that the upper limit of zero can be estimated as 3/n. Therefore, it seems that you ought to substitute 3 in the "0" cell but when I do that, I get an abnormally large confidence interval, larger than if I had done the original calculation using the calculator with a 1 in the box. Does anyone know what the correct approach to this is? Barney Eskin [log in to unmask]