Monitoring Aggregated Poisson Data for Processes with Time-Varying Sample Sizes
Open Access
- 1 July 2017
- journal article
- Published by Universidad Nacional de Colombia in Revista Colombiana de Estadística
- Vol. 40 (2) , 243-262
Abstract
This article deals with the effect of data aggregation, when Poisson processes with varying sample sizes, are monitored. These aggregation procedures are necessary or convenient in many applications, and can simplify monitoring processes. In health surveillance applications it is a common practice to aggregate the observations during a certain time period and monitor the processes at the end of it. Also, in this type of applications it is very frequent that the sample size vary over time, which makes that instead of monitor the mean of the processes, as would be in the case of Poisson observations with constant sample size, the occurrence rate of an adverse event is monitored.Two control charts for monitoring the count Poisson data with time-varying sample sizes are proposed by Shen et al. (2013) and Dong et al. (2008). We use the average run length (ARL) to compare the performance of these control charts when different levels of aggregation, two scenarios of generating of sample size and different out-of-control states are considered. Simulation studies show the effect of data aggregation in some situations, as well as those in which their use may be appropriate without significantly compromising the prompt detection of out-of-control signals. We also show the effect of data aggregation with an example of application in health surveillance.Keywords
This publication has 18 references indexed in Scilit:
- CUSUM procedures with probability control limits for monitoring processes with variable sample sizesIIE Transactions, 2016
- Weighted CUSUM Control Charts for Monitoring Poisson Processes with Varying Sample SizesJournal of Quality Technology, 2011
- The Role of Data Aggregation in Public Health and Food Safety Surveillance2010
- Surveillance Strategies for Detecting Changepoint in Incidence Rate Based on Exponentially Weighted Moving Average MethodsJournal of the American Statistical Association, 2008
- Should Observations Be Grouped for Effective Process Monitoring?Journal of Quality Technology, 2004
- Control Charts and the Efficient Allocation of Sampling ResourcesTechnometrics, 2004
- A general approach to modeling CUSUM charts for a proportionIIE Transactions, 2000
- Design of Optimal Exponential CUSUM Control ChartsJournal of Quality Technology, 1994
- Optimal surveillanceBiometrika, 1991
- Monitoring Poisson Observations Using Modified Exponentially Weighted Moving Average Control ChartsCommunications in Statistics - Simulation and Computation, 1990