Minimum distance estimation using the Lotka limit for age dependent branching processes (Q4378477)

A Bellman-Harris type age-dependent branching process \(Z(t)\) is considered with a parametrized parent lifetime distribution \(G(t,\theta)\) with \(G(0+, \theta)=0\) \(\forall \theta\) and an offspring probability distribution \((p_j)_{j= 0}^\infty\) with mean \(1<m< \infty\). A joint estimate of \((\theta,m)\) is obtained which requires only enumeration of all living members' ages of the population at the present time \(t\). The estimator (called Wolfowitz-Minimum Distance estimator) is constructed as minimum supremums norm estimator between the empirical age distribution \(A(x,t)=\)(number of individuals living at time \(t\) with age \(\leq x) \)/(total number of individuals living at time \(t)\) and the asymptotic Lotka limit distribution \[ A(x) =(m\alpha/(m-1)) \int^\infty_0 e^{-\alpha t} \bigl(1- G(t,\theta) \bigr)dt \] when \(\alpha= \alpha (m, \theta)\) denotes the Malthusian parameter of the Bellman-Harris process. Strong consistency is proved for a supercritical Bellman-Harris process with parameters in a compact set if the density \(g(t,\theta)\) of \(G(t,\theta)\) is subject to the Harris condition \(\int^\infty_0 |g(t, \theta) |^p dt< \infty\) for \(p>1\) and if the process is on the explosion set where \(Z(t)\) diverges for \(t\to\infty\). Identifiability conditions and suitable metrics on the parameters space and the Lotka limit space are established.