This paper proposes a semiparametric analysis of proportional hazards
models. This approach consists in specifying the relation between a du
ration and explanatory variables, without specifying the data distribu
tion. The parameters involved in this relation are then considered as
parameters of interest, and the data distribution is treated as a nuis
ance parameter. We propose a Bayesian estimation method, the principle
of which is to specify a prior distribution on the nuisance parameter
. We then obtain semiparametric estimators for the parameters of inter
est, by computing their posterior distribution, conditional on the dat
a and integrated with respect to the nuisance parameter.