|
|
شما اينجا هستيد: صفحه اصلي » تبادل نظر » تئوريك » روشهاي آماري در اپيدميولوژي
| ا22/4/1386 ا6:02 بعدازظهر |
|
|
دكتر محمد انصاري پور
|
In a 2*2 table which is beter to report, OR or ORss?n
when should we report ORss?
|
| ا23/4/1386 ا11:37 بعدازظهر |
|
|
irzamin
|
A 2*2 table will not generally recommend anything by itself regarding which effect measure to be used. Odds ratio is used as an 'effect measure' in a case-control study with RARE DISEASE assumption (of course it can be reported in cohorts too). If the disease is not rare, the odds ratio is NOT a good estimate of relative risk. About the bias introduced by small samples and ORss, there is paper to read (http://links.jstor.org/sici?sici=0006-341X(198606)42%3A2%3C351%3AOTBOCU%3E2.0.CO%3B2-3). If you do not have access to the full text of this paper, send it to my consultation section, and I 'll upload it for you |
| ا24/4/1386 ا9:02 صبح |
|
|
irzamin
|
I want to add one more comment on calculating the confidence interval for an odds ratio. Let's say we have a 2 x 2 table, and we got the estimate of the odds ratio. The asymptotic standard error of natural log of the odds ratio is: sqrt(1/a +1/b+1/c+1/d), where a, b, c and d are the counts of 2 x 2 table. Now we can calculate the confidence interval for the odds ratio, because we have its estimate and also its standard error. If we note the formula of the standard error we will see that any small count(s) in a cell(s) in 2 x 2 table will increase the standard error substantially. In the other words, smaller cell counts are associated with larger standard error and subsequently wider confidence interval for the odds ratio. So when the counts are very low in 2 x 2 table ORss should be preferable to OR, but consult with the paper I mentioned above |
| ا24/4/1386 ا11:01 بعدازظهر |
|
|
irzamin
|
In my discussion on calculation of the confidence interval (above), I meant calculation of the confidence interval for the NATURAL LOG of the odds ratio because the formula provides the asymptotic standard error of natural log of the odds ratio |
|
|
|
|
