One key challenge to the monitoring of inhomogenous Poisson processes with varying sample sizes is that sample size changes may mask incidence rate changes, leading to poor performance from the conventional Poisson CUSUM chart. Proposed is a class of weighted CUSUM schemes which apply weight functions to the likelihood ratio statistic to compensate for varying sample sizes. Comparisons to the traditional Poisson CUSUM method and to other proposed methods favor this proposal. An example in health care surveillance illustrates the proposed method.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados