An inference procedure for conflict models (Q2641051)

The authors consider a model for conflict between two forces with initial numbers \(n_ 1\) and \(n_ 2\) of combatants. The conflict is assumed to form a continuous time Markov process whose transition probabilities are specified and depend on two parameters \(\lambda_ 1,\lambda_ 2\). The paper is concerned with inference about the ratio \(\theta =\lambda_ 1/\lambda_ 2\), for which an estimator can be derived using martingales in continuous time. By equating a suitable martingale to its mean, the authors derive an estimator \({\tilde \theta}\) depending only on initial and final values of the numbers of combatants. Examples are given for the Linear Law Model, the Square Law Model, and the Suicide Attack Model. Some asymptotic results are obtained for \({\tilde \theta}\); the process is then generalized to the case where \(\lambda_ 1,\lambda_ 2\) are time dependent.