| Summary: | Occasionally, during the statistical process monitoring it is not possible to make measurements with precision on the observed units; therefore, it should be necessary to work with censored observations. Consequently, a direct application of the traditional Shewhart control charts would not be appropriate. Under these circumstances, it should be implemented control charts that include both the censoring involved and a suitable distribution for the resulting data. Under an inspection scheme, where no defective units are censored and a high censoring is presented on the stable state of the process, three control charts for X ̄ are proposed, in the case of monitoring interval-censored data. Each of them has a different way to impute the censored data. In the first one, each censored observation is replaced by its conditional expected value (CEV); in the second one, the censored data is replaced by the upper limit of the interval which the censored data belongs to, and in the third one, it is replaced by a random value of a uniform distribution within the censoring interval, considering a typical distribution and censoring type I. The power for detecting deviations from the mean for the proposed charts is compared by means of simulation with the Shewhart chart that ignores censoring.
|
|---|
