
Exposure–risk modeling for time-to-event data 155
has been well developed, hence the reader is referred to a standard text such
as Schoenfeld (1983). In general, sample size calculation for exposure–risk
modeling is rather complex, since the information matrix for model parameters
cannot be easily calculated. The following simulation tools provide an easy way
for applied statisticians and modelers. The basic idea is to use simulation in
combination with model fitting to estimate the information matrix with a large
sample simulation. Then the sample size needed to achieve a given prec ision
(e.g., given by a target SE) can be determined accordingly.