Parameter estimation causes a considerable stochastic error in standard Shewhart charts. This problem can be solved using suitable correction factors. But if the normality assumption itself fails, in addition a nonvanishing model error will occur. By then, a nonparametric alternative, such as the recently proposed MIN chart, might be a better idea throughout. However, for those reluctant to give up on the Shewhart chart, a third possibility is offered here. One sticks with this traditional chart as long as the data suggest that the resulting model error is acceptable. Only if this not the case, the MIN chart kicks in and as such serves as a nonparametric safeguard.
|Name||Memorandum / Department of Applied Mathematics|
|Publisher||Department of Applied Mathematics, University of Twente|