[R-meta] Need suggestion on forest plot results

Guido Schwarzer @c @end|ng |rom |mb|@un|-|re|burg@de
Wed Jun 3 13:59:38 CEST 2020


(I haven't noticed that I got a personal email by Awais - find below my 
reply)

Awais,

As expected this is a rounding problem and you can do nothing about it 
but get more precise estimates (which are probably not available).

The distances between lower and upper CI limit to the log hazard ratio 
are different (due to rounding to two digits):
lnHR = log(1.93)
lowHR = log(1.23)
uppHR = log(3.00)
lnHR - lowHR
uppHR - lnHR

The standard error is approximated from half the distance between lower 
and upper 95% CI:
(uppHR - lowHR) / (2 * qnorm(0.975))
meta::metagen(lnHR, lower = lowHR, upper = uppHR)$seTE

Which results in the minor shift of the confidence limits:
meta::metagen(lnHR, lower = lowHR, upper = uppHR, sm = "HR")

Best wishes, Guido



More information about the R-sig-meta-analysis mailing list