Eliminating the influence of serial correlation on statistical process control charts using trend free pre-whitening (TFPW) method

Abstract

A key assumption in traditional statistical process control (SPC) technique is based on the requirement that observations or time series data are normally and independently distributed. The presences of a serial autocorrelation results in a number of problems, including an increase in the type I error rate and thereby increase the expected number of false alarm in the process observation. However, the independency assumption is often violated in practice due to the influence of serial correlation in the observation. Therefore, the aim of this paper is to demonstrate with the hospital admission data, the influence of serial correlation on the statistical control charts. The trend free pre-whitening (TFPW) method has been used and applied as an alternative method to obtain residuals series which are statistically uncorrelated to each other. In this study, a data set of daily hospital admission for respiratory and cardiovascular diseases was used from the period of 1 January 2009 to 31 December 2009 (365 days). Result showed that TFPW method is an easy and useful method in removing the influence of serial correlation from the hospital admission data. It can be concluded that statistical control chart based on residual series perform better compared to original hospital admission series which influenced by the effects of serial correlation data.

Cross-correlating deforestation and atmospheric indexes to local rainfall and temperature patterns through detrended multiple cross-correlation analysis

“…the annual cycle) were removed from each time series. Serial autocorrelation usually generates spurious results as stated by Desa and Jemain (2014). Thus, we reduced the serial autocorrelation using the von Storch (1995) method.…”