Articles - Manufacturing Quality Control Software - Because your profit is on the bottom line...

Articles

Overcoming the Statistical Avalanche

With statisticians and quality engineers required to analyze ever-increasing amounts of complex quality and business data, the need for statistical software is certain.

Ten of the world’s leading statisticians addressed the future of industrial statistics in the latest issue of Technometrics, a trade journal focusing on the field of statistics. Though it’s rare for experts to agree, this panel unanimously predicts the major challenge for future statisticians and quality engineers will be the massive volume of data with which they are asked to work.

“The new statistical problems posed by large data sets will be a huge challenge in creating an analog of Six Sigma … that allows ordinary mortal industrial statisticians to work in a disciplined fashion in these situations,” explained Nicholas Fisher, a statistical consultant and researcher for more than 30 years. “In other words,” Fisher continues, “how do we convert the art of analyzing massive data sets into a science, available to the masses?” (Fisher is the founder of ValueMetrics Australia, a consulting and research firm specializing in performance measurement and a visiting professor of statistics at the University of Sydney.)

Fisher thinks the industrial statistician’s traditional role in quality programs such as Six Sigma is no longer suited to the challenge of analyzing enormous amounts of business and industrial data. “Any statistical process control involved will almost certainly be far ahead of current practice because of the size and nature of the data. Yet the clients' basic quality requirements for timely and actionable results endure.”

The primary, if not the only, solution to the challenge of ever-increasing amounts of statistical data, the panelists concur, is statistical software.

Jon Kettenring, a statistician for industrial-telecommunications research organizations for nearly 40 years, is involved in a unique program for retired industrial scientists at the Research Institute for Scientists Emeriti (RISE) at Drew University in Madison, New Jersey. He feels the number one priority for the next generation of statistical software is “genuine” ease of use.

Soren Bisgaard, a professor at the University of Massachusetts-Amherst and at the University of Amsterdam, thinks statistical software is essential because of the need “to create better statistical methods, especially more intuitive and easier to understand multivariate time series techniques, software combined with statistical graphics to enable the user to understand implications of the often complex multivariate statistical analysis and modeling.”

Another driver for expanded reliance upon statistical software is the growing use of complex, advanced statistical modeling in life sciences and other manufacturing industries. “Biopharmaceuticals is an especially interesting area of application for advanced industrial statistics applications," Bisgaard explains. “Because of their biological base, such processes exhibit large and difficult-to-control variability and often are extremely nonrobust.” He sees design of experiments, response surface methods, evolutionary operation (Box 1957) and multivariate process control for autocorrelated data as “indispensable tools for achieving high, reliable and uniform yields, as well as for maintaining high product safety standards.”

Bisgaard expects such statistical modeling to become more prominent across disciplines in industrial statistics and applications. “We usually think of statistical process control as useful for the control of manufacturing processes. But we ought to think of it more broadly as systems monitoring and control, where the system could be almost anything.” As an example, Bisgaard notes that automobiles are required by federal law, for environmental reasons, to be equipped with an onboard diagnostics computer to monitor a large array of powertrain performances.

The entire Technometrics article, titled The Future of Industrial Statistics: A Panel Discussion, can be viewed at the following web address: http://www.asq.org/pub/techno/index.html.
 
Quality Control Products SPC Software Applications Why Use WinSPC? SPC Information