Using Fisher scoring to fit extended Poisson process models (Q1775972)

The paper deals with the so called ``extended Poisson process model'' (EPPM) which is defined by \(\{Y_i(t),i=1,\dots,m\}\), where \(Y_i(t)\) are Markov processes describing the number of events occurring in the time interval \((0,t)\) and satisfying \[ P\{Y_i(t+\delta t)=n+1\mid Y_i(t)=n\}=\lambda_{in}\delta t +o(t),\quad P\{Y_i(t+\delta t)=n\mid Y_i(t)=n\}=1- \lambda_{in} \delta t +o(t). \] The authors focus on a special parametric form for the rate parameters \(\lambda_{in}\), i.e., they put \(\lambda_{in}=a_i(b+n)^c\), \(a_i>0, b>0, c<1\), and develop a Fisher scoring algorithm for the maximum likelihood estimation of the parameters for the mentioned EPPM with covariate dependent means. An application of such algorithm to analysis of school absence data is discussed.