Screening Outliers in Normal Process Control Data with Uniform Residuals

The use of sequential uniform residuals is proposed to screen outliers in process control data. The exact distribution theory of these statistics allows a precise control of the number of observations incorrectly rejected when the process is in control and generated by a normal process distribution. It is also shown that the technique given here has attractive inference properties for detecting shifts in the process distribution mean. The power for detecting a mean shift of specified magnitude is given in terms of a noncentral Student t distribution function, and it is observed that a central Student t distribution gives a good approximation to this power for most cases of interest.