Optimal estimation for the functional Cox model (Q309740)

The authors consider the functional Cox model with right-censored data and with vector covariate \(Z\) and functional covariate \(X(t)\) that represents the hazard function by \(h(t| X)=h_0(t)\exp\{\theta'_0 Z+\int_{S}X(s)\beta_0(s)ds\}\), where \(h_0\) is an unspecified baseline hazard function, \(\theta_0\in \mathbb R^{p}\) is an unknown parameter, \(\beta_0\) is an unknown coefficient function. It is studied the convergence of the estimator \((\hat\theta,\hat\beta)\) under the framework of the reproducing kernel Hilbert space. The authors obtain the estimator by maximizing a penalized partial likelihood and establish \(\sqrt{n}\)-consistency, asymptotic normality and semiparametric efficiency of the estimator \(\hat\theta\) of the finite-dimensional regression parameter. It is proved that the estimator of the coefficient function achieves the minimax optimal rate of convergence under the weighted \(L_2\)-risk. An efficient algorithm is developed to estimate the coefficient function. Implementation of the estimation approach, selection of the smoothing parameter and calculation of the information bound \(I(\theta)\) are discussed. Numerical studies are presented.