No label defined (Q4264696)

Assume that under the (column) vector of covariates, \(x:[0,\infty)\mapsto B\subset {\mathbb R}^m\), a life time \(T_{x(\cdot)}\) is a non-negative random variable with survival function \(S_{x(\cdot)}(t) = P(T_{x(\cdot)}>t)\), \(t\geq 0\). The authors consider the generalized additive-multiplicative model \[ (*)\;\;S_{x(\cdot)}(t) = G(\int_0^tr(x(\tau)) dH(S_0(\tau))+\int_0^ta(x(\tau))d\tau). \] Here \(G\) is some specified survival function, \(H=G^{-1}\), \(S_0\) is an unknown baseline survival function, while \(r(x(\cdot))\) and \(a(x(\cdot))\) are parametrized as \[ r(x(\cdot))=\exp({\beta}^Tz(\cdot)),\quad a(x(\cdot))={\gamma}^Tw(\cdot), \] where \(\beta=(\beta_1,\ldots,\beta_p)^T\) and \(\gamma=(\gamma_1,\ldots,\gamma_s)^T\) are vectors of unknown regression parameters, \(x^T(\cdot)=(z^T(\cdot),w^T(\cdot))\), with \(p+s=m\). (If \(p=m\) or \(s=m\), then \(a(x(\cdot))\equiv 0\) or \(r(x(\cdot))\equiv 1\), respectively.) Model \((*)\) contains most of the well-known regression models in survival analysis. Estimators for \(\beta, \gamma\) and \(S_0\) are constructed in the case where \(x(\cdot)\) is a predictible random process, and their asymptotic properties are studied.