| Popis: |
On control chart, one may use regression, cusum and exponential smoothing methods to detect any change in the trend of the process. When the data is observed with equally spaced time interval we show how to obtain the estimates of the trend for each method. Three different estimation methods are compared through simulation in terms of efficiency, MSE, MAPE and MAD. It is observed that OLS performs well in general. Cusum is found to be most sensitive to the outliers even though it is simple to compute. We recommend the use of the EWMA method when the error distribution is heavy tailed. |
