10 Statistical Concepts Everyone Should Know

10 Statistical Concepts Everyone Should Know

By Discovery Lean Six Sigma

0/5 stars (0 votes)

10 Statistical Concepts Everyone Should Know

Bill Kappele, technical director and instructor at consulting company Objective DOE, does a great job of teaching individuals and organizations how to use experiments to reach a powerful level of improvement knowledge. He’s compiled a list of 10 statistical concepts that every engineer should know.

Bill’s list is excellent. In fact, I believe his 10 key concepts should be known by everyone tasked with making process or system improvements. But because Bill’s audience is primarily engineers, the language he’s used may be hard for some to get through. Here’s how I would tweak Bill’s list to extend it’s value to a larger audience:

  1. The performance of every process or system contains variation. This variation is a fact of life. Sometimes the variation is small—sometimes it is large.
  2. Even the method and equipment you use to measure performance—called your “measurement system”—adds extra variation to performance data you collect. One way to know how much your measurement system is contributing to your view of the the variation in your performance data is to make a few repeated measurements on the same part or output. Differences in your repeated measurements show you the size of your measurement variation.
  3. The variation in a measurement system can be large enough to make it inadequate for its intended use. Even expensive systems in good working order may not provide accurate enough or precise enough measurements for your situation.
  4. You need to perform what’s called a “measurement system analysis” to determine if your measurement system is adequate for its assigned or intended use. If you don’t perform this study, or have someone perform it for you, you might be acting falsely on misleading information from your measurement system.
  5. Valid statistical predictions rely on the assumption that your process or system will be configured and operated tomorrow the same way that it is configured and operated today. Like a moving target, inconsistencies in configuration or operation negate the predictive power of statistics.
  6. Consistent process/system behavior is called “stable” or “in control” behavior. You can use statistics to make valid predictions about future performance only if your process or system is stable or in control.
  7. “Statistical process control,” or SPC for short, helps you determine if your process or system is stable or in control.
  8. Input factors to a process or system, like temperature or time, can interact—meaning that together, these input factors will have a diffrent effect on the output performance of the system or process than when they act alone.
  9. Nature often works through factor interactions. Knowing the effect each factor has when acting by itself does not tell the whole story. You also need to know how the factors interact.
  10. “Design of experiments,” or DOE for short—an approach where multiple experimental factors are changed between each experiment run—detects and quantifies factor interactions. Classical change-one-factor-at-a-time experiment approaches cannot.





Original: http://todayssixsigma.com/2008/10-statistical-concepts-everyone-should-know/
By: admin
Posted: April 30, 2008, 7:03 am

comments powered by Disqus

Discovery Lean Six Sigma

Dummy user for scooping articles

I'm a dummy user created for scooping  great articles in the network for the community.