An Examination of the Robustness to Non Normality of the EWMA Control Charts for the Dispersion

Maravelakis, Petros E.; Panaretos, John; Psarakis, Stelios

doi:10.1080/03610910500308719

Cited by 52 publications

(29 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This result agrees with Montogomery 1 who states for the EWMA "It is almost a perfectly non-parametric (distribution free) procedure". Maravelakis et al 47 study the robustness to normality of the EWMA by tabulating characteristics of the run length distributions (e.g. ARL) for observations generated by several gamma distributions.…”

Section: The Modified Ewma Control Chart For Correlated Datamentioning

confidence: 99%

Multivariate Control Charts Based on Bayesian State Space Models

Triantafyllopoulos

2006

Quality & Reliability Eng

View full text Add to dashboard Cite

This paper develops a new multivariate control charting method for vector autocorrelated and serially correlated processes. The main idea is to propose a Bayesian multivariate local level model, which is a generalization of the Shewhart-Deming model for autocorrelated processes, in order to provide the predictive error distribution of the process and then to apply a univariate modified EWMA control chart to the logarithm of the Bayes' factors of the predictive error density versus the target error density. The resulting chart is proposed as capable to deal with both the non-normality and the autocorrelation structure of the log Bayes' factors. The new control charting scheme is general in application and it has the advantage to control simultaneously not only the process mean vector and the dispersion covariance matrix, but also the entire target distribution of the process. Two examples of London metal exchange data and of production time series data illustrate the capabilities of the new control chart.

show abstract

Section: The Modified Ewma Control Chart For Correlated Datamentioning

confidence: 99%

Multivariate Control Charts Based on Bayesian State Space Models

Triantafyllopoulos

2006

Quality & Reliability Eng

View full text Add to dashboard Cite

show abstract

“…It is probably used more often in the literature because it is conceptually easy to understand. Finding the in-control ARL or OOC ARL actually determines the starting point at which practitioners begin to count the number of plotted observations (Maravelakis et al, 2005).…”

Section: Performance Assessment and Comparison Of Methodsmentioning

confidence: 99%

“…Control charts are designed to serve this purpose by focusing on two strategies. Firstly, we want the chart to signal false alarm (false alarm occurs when a series signals before the end of its in-control readings) as planned when we are in control and secondly, when the process is out-of-control (OOC), we want the control chart to signal as soon as possible (Maravelakis et al, 2005). With reference to these two strategies, the most popular technique for evaluating the performance of a control chart is the average run length (ARL), which is based on the run length distribution (Jensen et al, 2006;Maravelakis et al, 2005;McCracken and Chakraborti, 2013).…”

Section: Introductionmentioning

confidence: 99%

Non-parametric change-point approach for monitoring shifts in process location and variability

Afolabi

Osanaiye

Akpa

2015

IJQET

View full text Add to dashboard Cite

Abstract:In statistical process control, detecting if the process is in control and the position of shift in an out-of-control process are critical research problems. If the normality assumption is satisfied, work has advanced in detecting shifts in mean and/or variance. However, the normality assumption is often not satisfied in many real life situations. We suggest a non-parametric Lepage-type change-point (LCP) control chart for jointly detecting process shifts in mean and variance, under non-normality. A comparison between our proposed method and a generalised likelihood ratio (GLR)-based method was made. Process data were simulated following normal and Laplace distributions. The performances of LCP and GLR were assessed and presented, using evaluated average run lengths, under the distributions considered. The LCP competed favourably with the GLR in a normal distribution. However, LCP outperformed GLR under the heavy-tailed distribution considered. We recommend the new approach for short-run situations where the underlying distributions are usually unknown.

show abstract

“…An alternative measure that is more reliable and acceptable is the median of the run length distribution (MDRL) (Abbasi & Miller, 2013;Maravelakis et al, 2005). Usually the standard deviation of the run length distribution (SDRL) is estimated whenever the average run length is calculated (Abbasi & Miller, 2013;Maravelakis et al, 2005). In this article we used these three measures i.e.…”

Section: Performance Evaluationmentioning

confidence: 99%

“…F. Gan, 1993;Woodall, 1983) as it produces the misleading results when the run length distribution is skewed or if it follows a non-normal distribution or the form of the distribution is unknown. An alternative measure that is more reliable and acceptable is the median of the run length distribution (MDRL) (Abbasi & Miller, 2013;Maravelakis et al, 2005). Usually the standard deviation of the run length distribution (SDRL) is estimated whenever the average run length is calculated (Abbasi & Miller, 2013;Maravelakis et al, 2005).…”

Section: Performance Evaluationmentioning

confidence: 99%