Regression Parameter Estimation from Panel Counts (Q4455931)

A counting process \(N_i(t)\) associated with subject \(i\) (\(i=1,\dots,n\)) is observed only at a finite number of time points \(\xi_{il}, l=1,\dots,K_i\), (e.g., \(N_i(\xi_{il})\) can be the number of bladder cancer tumours observed for the \(i\)-th patient at \(l\)-th visit to a clinic). \(Z_i\) is a \(p\)-dimensional covariate. It is assumed that \(E(N_i(t)\mid Z_i=z_i)= \Lambda_0(t)\exp(\beta^t z_i)\), where \(\Lambda_0\) and \(\beta\) are unknown. The authors propose two estimators for \(\beta\) based on the estimating equation technique. The asymptotic normality of the estimators is demonstrated and estimates for their covariance matrices are proposed. Results of simulation and applications to bladder cancer data are presented.