Process Variance Analysis For When Data Are Autocorrelated

A primary assumption of many procedures in statistical process control and in process capability analysis is that the observations taken from the process are independent. However many processes exhibit a certain degree of autocorrelation. In this paper we discuss the effects that autocorrelation may have on process variance and capability indices and we show that when autocorrelation exists, variance estimate ceases to be unbiased. Also two simple approaches are developed to calculate process capability indices in the presence of autocorrelation. The first approach is based on the assumption of rational subgroups of means and variance analysis. The second approach is based on the calculating process capability from an autoregressive model.