dear all,
clayton and kaldor outline a way to calculate empirical bayes estimates
(ebe) to replace conventional standardized mortality ratios (smr) in
mortality rate analysis (biometrics 43, 671-681, 1987).
in essence, if smr=o/e, then ebe=(o+nu)/(e+alpha), where the smoothing
factors nu and alpha are calculated
from the body of data normally used for the calculation of smr's.
my query: how is the variance of the ebe's calculated? is there a
straightforward formula just like there is for smr, ie var(smr)=o/e^2, or
var(log(smr))=1/o ?
any formulas if you know the immediate answer, or pointers thereto would be
greatly appreciated.
thanks a lot in advance
laszlo
(note: right now i don not have access to the above cited article, i know
about it from references. i am not even sure it gives the variance formula
explicitly. i will of course get it if you name it as the best source.)
|