# 1.5 Sigma Shift Not A quot;Standardquot;

Six Sigma – iSixSigma Forums Old Forums General 1.5 Sigma Shift Not A quot;Standardquot;

Viewing 8 posts - 1 through 8 (of 8 total)
• Author
Posts
• #27144

Joe Perito
Participant

Ken Meyers, sorry for the late reply. I take liberties in the use of the term “sigma”. I do this in relation to the logic of the amount of variation and the published table that everyone knows, “The areas on the normal bell curve”. This table is used both for data for individuals, and for sbubgroups of data. The area under the normal curve for a Z factor of 3.00 (or +-/ 3 sigma) is 99.73%. This is true for +/- 3 sigma for individuals or for +/- 3 standard errors of the mean… although you will never of heard anyone other than me (just now) use “standard errors of the mean” in relation to the table. The common language is “confidence levels”, Z factors, or most commonly “sigmas”… although I admit this is not explicitly correct terminology if you are refering to data based on averages. Still the table is applicable to control charts and frequency distribution based individuals and averages. The calculations for individual’s variation and for average’s variation are different, the confidence levels, the Z factors, and the area under the normal curve are the same. The amount of process shift that could potentially take place is NOT a standard number. Hence, the logic of Schewhart control charts and “assignable causes”.

0
#66101

Cone
Participant

Joe,

I ask you once again — what is your experience with the process shift — what are you finding?

I ask this because I think there is a very informative discussion that follows on improvement methods and results when the shift is small vs when the shift is big.

Let’s take all the theoretical stuff to a practical level where folks can actually learn from all the banter.

Do you know how many teeth an ox has?

Gary

0
#66102

Ken Myers
Participant

Joe,

Despite the muddle, there is nothing you have said that is new or unique to this discussion. My point in the previous message to you is that you can alter the shift in the mean of the process using a control chart of averages. I’ve done this in the past with processes that I desired tighter control on. If I wanted to control a process to less than +/-1.5 standard errors I would use a sample size per sub-group larger than 4. This means the some of the variation I once called “common cause”, would now be classified as “special cause”. As such, I would have additional improvement work to conduct. As Gary aptly said sometime earlier that work would entail variation reduction. Certainly, this make some sense to you…

Gary, if you’re reading this you’ve peaked my interest! So, how many teeth does an ox have? Joe, feel free to jump in anytime…

Regards,

Ken

0
#66104

Cone
Participant

Ken,

I don’t know how many teeth an ox has. We have two choices —

1) We can sit and debate our opinions about how many, or

2) We can get off our philosophical high horses and go where the ox lives and get some data.

Note: It is reported that the ancient Greeks used to sit in the town center and debate such heady topics. They did not make the walk out to the country to find the real answer. Not much has changed.

I think the ox has 5 teeth, what do you think?

0
#66107

Ken Myers
Participant

Gary,

I think this forum makes it difficult to do much more than have healthy discussions about the topics presented.

I think you first have to see new concepts in your mind before you can act.

I think there is a difference between a philosophical point, and a technical point.

I think the cause of much confusion and argument in with any point of disscussion is due to language.

I think clarity of thought brings us closer to action.

I think learning is also about listening with empathy.

I think an ox has 26 teeth… What do you think?

0
#66109

Joe Perito
Participant

Ken, I wonder if we have a diffinition problem. You can not make a “process mean” change by changing the the distance between control limits. The precess mean (or average) will always be the same regardless of whether you calculate the average off of the individuals, or the averages… regardless of what the subgroup sizes are. Xbar of the individualls will always equal double Xbar The grand mean of the means). Likewise, you can not arbtrarily change the subgroup size on a set of data previously compiled from subgroups of a different size… because the samples were not sampled in the manner that you want to change the new subgroup size to. A process mean change (In my terminology) requires a “process” adjustment to shift its set point.

0
#66111

Cone
Participant

Joe,

Enough with the stats lectures. What is your actual experience? My experience says that organizations don’t behave like SPC rules which is why all the bantering to see who knows the most about stats is meaningless.

Tell us what you are actually experienceing Joe. Experience is so much richer that theoretical knowledge. If you tell us everything you touch behaves like a control chart, everyone will know you are blowing smoke.

Lets talk reality, forget the stats lectures. We can get those for a dime a dozen at the local college.

Gary

0
#66112

Cone
Participant

Joe,

How about weighing in on the process shift discussion.

Gary

0
Viewing 8 posts - 1 through 8 (of 8 total)

The forum ‘General’ is closed to new topics and replies.