ASQ - Team and Workplace Excellence Forum


February 1998

Articles

Business And Sports, One-On-One

Measurement On The High Seas

Scientists Develop Formula For Multinational Teamwork

Part-Time Statistics For Full-Time Results

Volunteers Wanted: Must Be Team Player, Success Minded



Columns

Chasing Good Examples
by Peter Block

Individual Change Key To Org. Change
by Cathy Kramer


Features

Brief Cases
Business News Briefs

Views for a Change

Pageturners
Book Review

 

Part-Time Statistics For
Full-Time Results
A Basic Understanding of Statistics Benefits any Profession Organization

Many people perceive statistics as being difficult or complicated. Ask Dr. Neil Poulsen. "I collect evidence to reach this conclusion every time someone asks me what I do for a living. As soon as I tell them I'm a statistician their eyes glaze over, they rock back on their heels a bit as they recall that one statistics course they took in college. 'Oh no, I took that class and barely escaped with a C!' "When people think of statistics, they often think of complex methodologies that seemingly place statistics beyond the reach of everyone except those who, like statisticians, have made an extensive study of this science," Poulsen noted with some resignation. "The perception is that a company cannot have a statistics program unless it has a statistician. This is especially unfortunate for smaller companies which may not be able to afford a full-time statistician."

But this need not be the case. Poulsen, a statistician in the semi-conductor industry with Intel Corporation, believes the work of both W. Edwards Deming and Kaoru Ishikawa suggests otherwise. He recommends quality facilitators familiarize themselves with the work of these two authors - work that emphasizes the importance of basic statistical methods fundamental to quality improvement - and suggests how a small company might develop a statistics effort without hiring a full-time statistician.

How to Rethink Statistics
Deming introduced two types of studies, enumerative and analytic, in a way that fundamentally challenged the way we think about and practice statistics. Before defining enumerative and analytic studies, quality facilitators need to understand some basics about how data samples are collected.

A manufacturing facility has an injection molding machine which produces plastic screws. If the purpose of the data sample is to evaluate the strength of plastic screws, then the population would be all screws produced in the past and future under similar conditions of temperature, pressure, etc. In this instance, we have no choice but to use the concept of a frame, an intermediate collection of screws from which to select a random sample. Otherwise, half the sample would need to be selected from the future, which is not possible. In this case, we might define the frame to be a lot of screws, produced on a given day or morning.

If the purpose is only to determine whether or not to ship that or some other particular lot of screws and not to estimate the strength of screws produced overall by the injection molding machine, then the frame itself (the lot) is the population about which we want to make inferences.

This example distinguishes the difference between enumerative and analytic studies. Enumerative studies are those for which we are interested in drawing inferences only about the frame, not in the process that generates the frame. Potential action will be taken on the frame. On the other hand, analytic studies are those in which our interest is in the process that produces the frame where action will be taken on the process. Poulsen lists the characteristics of each type of study:
Enumerative Studies (Interested in frame)
o Fall under the category of inspection, which may or may not add value
o Focus on the past
o Address symptoms without necessarily addressing root causes
o Determine all there is to know about the population with 100% sampling
Analytic Studies (Interested in process)
o Involve drawing conclusions about how to improve a process
o Focus on the future
o Address root-cause solution of problems
o Not possible to conduct 100% sampling

"Deming made it very clear that data is value-added only to the extent that it allows us, through theory, to predict and draw conclusions about the future," noted Poulsen.

Keep it Simple
Graphical and more empirical and descriptive statistics have moved to the forefront of statistical application. Ishikawa observed "Introductory statistics… if used skillfully, will enable 95 percent of the workplace problems to be solved. In other words, intermediate and advanced statistical tools are needed in about only 5 percent of the cases." This supports the rationale that a small company can have an effective statistics program without hiring a statistician. Basic statistical methods are intuitive and easy enough to be applied and understood.

The majority of activities of an internal consulting statistician within a company usually fall within one of three categories: provide statistical training, provide statistical consulting or drive statistically related projects. The statistician also should help people with problems relating to data collected and how best to evaluate that data. For more difficult statistical problems that might occur, the company can hire an outside statistical consultant who can be called upon to provide assistance and guidance.

"Let's dispel that bugaboo that statistics is difficult or complicated," implored Poulsen. "Basic statistical methods and effective statistical programs are well within the reach of small companies. An organization should be intrigued enough by the power and benefits gained from basic statistics to undertake and successfully implement a statistical program where they might not otherwise have made the attempt."

February '98 News for a Change | Email Editor

 



  • Print this page
  • Save this page

Average Rating

Rating

Out of 0 Ratings
Rate this item

View comments
Add comments
Comments FAQ

ASQ News