Twenty Nickels
If you save twenty nickels, you’ve made a dollar
by
Raymond Matkowsky
www.datastats.com
email: rdm@datastats.com
**********************************************************************
Issue 9-14 September 2014
AOV – Not As Easy As You Might Be Led To Believe!
by
Raymond Matkowsky
Over the past several years I have read numerous articles that leads one to believe that all they have to do is place a disc in a computer, plug in the pertinent data, and you retrieve a complete picture of your system and the how to improve it. Nothing can be further from the truth and it annoys me to read about this miraculous procedure. An analysis of variance is a powerful technique in the right hands.
First of all, I have been doing analysis of variance procedures for over forty years, long before it became a popular method. One of my early employers sent me to school all day, once a week for six months to learn how to do it. It is not something you learn overnight as many of these vendors will let you believe.. After all this time, I still believe that I do not know everything that there is to learn.
First of all, you need a solid background in the area you are going to study, its manufacturing, and associated costs. You need to determine what your goals are. Only then can you pick the factors that you are going to study. The more factors you pick, the greater the potential error. I generally do not recommend more than three at a time. I would also suggest that you run for several months using the outcome of those factors before proceeding to others.
There may be interactions that are responsible for the outcomes. Many are not even on the horizon. Some factors may be completely discounted. A serious mistake! You cannot overlook these. I had one client that was running the same process for years not knowing that much of his variability and seconds production was due to his catalyst and process temperature. He discounted the catalyst for a long time. Other factors are equipment, stirring rates, heating rates, order of addition, and don’t forget the human factor. The way your workers carry out the process can certainly effect it. It does take a bit of intuition at times and a great deal of interpretation.
Many times something will work splendidly in the lab and be an utter failure in production. It is recommended that one start with lab results that go into a pilot plant and the pilot plant modifications go to a production phase. This may sound time consuming, but in the long run it may actually be a shorter method.
It is not always in your best interest to take factors off the top of the menu of results. Sometimes it is much better to choose a combination of lower level factors. Again interpretation!
In short, it takes a healthy knowledge to run anything but the most simplistic analysis and if you are going to run a simple analysis there are less time consuming methods that will give you an adequate result. I cringe when someone has tried a do it yourself approached only to get poor results that are blamed on the technique.
************************************************************************
A Call For Reader Input
We at Data Stats would like to see this newsletter become a reader supported forum for help questions, answers, or general comments on anything appearing in this newsletter or Data Stats’ website. If you have a question, answer, or comment to contribute send them to me at newsletter@datastats.com. I will try to publish it here.
If you have an urgent question to ask, you don't have to wait for our newsletter to come out. We will try to find you a reader or one of our experts that may be able to help you. So, if you have a question, comment, or think that you can be of help, send us an email at once to: newsletter@datastats.com.
Also, in your email, please let us know if we have permission to share your email address with experts that may be able to exchange ideas with you directly.
Raymond D. Matkowsky
Copyright © 2014 Raymond D. Matkowsky
Data Stats
P.O Box 672
Old Bridge, New Jersey 08857-0672
What You May of Missed in Previous Issues
|
