Fitting the extended Poisson process model to grouped binary data (Q2488427)

Data of the form \(Y=(Y_1(1),\dots,Y_m(1))\) are considered, where \(Y_i(t)\) are independent Markov birth processes with intensity of the birth in the state \(Y_i(t)=n\), \(\lambda_{in}=a_ie^{\gamma\sqrt{n}}(N_i-n)\), where \(a_i\) and \(\gamma\) are unknown, and \(N_i\) are known. The means \(\mu_i=\mathbf E Y_i(1)\) are linked to the covariate vectors \(x_i\) by the logistic link \[ \log\left(\mu_i/(N_i-\mu_i)\right)=x_i'\beta. \] The authors propose an iterative re-weighted least squares algorithm for the calculation of the maximum likelihood estimate for \(\gamma\) and \(\beta\) by \(Y\). Analysis of zooplankton data is considered as an example.