Simulating Stochastic Populations. Direct Averaging Methods

Abstract

A method of directly computing the average behavior of stochastic populations is established, which obviates the time-consuming process of generating detailed sample paths. The method relies on suitably discretized time intervals in which nonlinearities are quasi-linearized to produce random variables with known expectations and variances. The pair of equations is directly solved to obtain the average behavior of the system at the end of a time interval based on its knowledge at the beginning of the interval. The sample path requirement for this process is considerably lower than that for the process over the entire simulation period. The efficiency of the method is demonstrated on the transfer of antibiotics resistance between two bacterial species which is a problem of mounting concern in fighting disease.