Exponentially Weighted Moving Average (EWMA) Control Charts for Monitoring an Analytical Process
Polona K. Carson* and Arthur B. Yeh
Department of Applied Statistics and Operations Research, Bowling Green State UniVersity,
Bowling Green, Ohio 43403
The exponentially weighted moving average (EWMA) control chart is very effective in detecting small shifts in process mean or variance, but so far has not been well presented in the field of analytical chemistry. The main difference from the Shewhart chart is that the EWMA chart combines current data with historical observations by essentially taking a weighted average with weighting factor w of the most current sample observations and historical observations. We show that the EWMA chart with 0.05 < w < 0.20 is more effective in detecting small shifts in mean and variance than the Shewhart chart. In addition, the EWMA chart can also be used to forecast the observation in the next period, which can help analysts take preventive actions before process departures to the out-of-control state. Another advantage of using the EWMA chart is its good performance for observations that are not normally distributed or are autocorrelated.