Smooth relative risks from a set of expected and observed number of cases
using a log-Normal model as proposed by Clayton and Kaldor (1987).
There are estimated by
\(\tilde{\beta}_i =\log((O_i+1/2)/E_i)\)
in order to prevent taking the logarithm of zero.
If this case, the log-relative risks are assumed be independant and to have a
normal distribution with mean \(\varphi\) and variance
\(\sigma^2\). Clayton y Kaldor (1987) use the EM algorithm to
develop estimates of these two parameters which are used to compute the
Empirical Bayes estimate of \(b_i\). The formula is not listed here, but
it can be consulted in Clayton and Kaldor (1987).