# Is a stable process also normal at the same time?

#### Jafri

##### Involved In Discussions
What I understand is that a stable process is one whose Xbar-R control chart is within spec. In other words, no special cause present.
But can we automatically assume from such a stable process that its output will follow normal distribution as well?

Is stability and normality one and the same?

#### AMIT BALLAL

Trusted
What I understand is that a stable process is one whose Xbar-R control chart is within spec. In other words, no special cause present.
But can we automatically assume from such a stable process that its output will follow normal distribution as well?

Is stability and normality one and the same?
Xbar-R chart should be compared with control limits, not specification limits. I hope, you wanted to say control limits.
If we use a variable control chart and if there is no special cause, there is a possibility of the process to follow normal distribution. But we cannot be 100% sure that it distributes normally, since we don't always use all rules i.e. to be applied to control charts. Hence normality and stability can be same and cannot. Hence better to check normality of your process (data).

You are saying the process is stable and whether it can be said that output will follow normal distribution as well?
In Xbar-R chart / any variable control chart, we take samples (which is again output of the process) and based on these we are studying behavior of the process. So, we are always speaking about normality of the process by studying output of the process (products).
Again, normality in case of SPC helps us to predict about the population using indices such as Cp/Cpk Or Pp/Ppk. Unless the process is normally distributed, we cannot properly predict about the population.

#### Bev D

##### Heretical Statistician
Staff member
Super Moderator
A process that is in statistical control DOES NOT necessarily have a Normal distribution. Nor is a Normally distributed process necessarily ‘in statistical control’. This is just a Zombie theory. It has been disproven and explained countless times since the 1930s when Shewhart first described the method of statistical control.

And yes we can make predictions about processes that are not Normally distributed...

Statistical control is based on the homogeneity of the process - as rationally subgrouped - and not on it’s distributional shape. A process in statistical control has random variation within predictable limits and exhibits no discernible patterns (beyond those that are controlled for by rational sub grouping).

PLEASE read the works of Donald Wheeler if you are interested in statistical process control. His articles are available for free on the Quality Digest web site. They are short and easy to read. He stays true to Shewhart who ‘invented’ SPC and to Deming and others who understood and promoted Shewhart’s work.
A great start is Wheeler’s article entitled “Probability Models do not generate your data”
Probability Models Don’t Generate Your Data | Quality Digest

#### Mike S.

##### An Early 'Cover'
Trusted
Bev nailed it. A "thanks" just didn't seem like enough.

#### AMIT BALLAL

Trusted
Bev nailed it. A "thanks" just didn't seem like enough.
Yes. He always does.

Sent from my CP8676_I02 using Tapatalk

Last edited:

#### Jafri

##### Involved In Discussions
PLEASE read the works of Donald Wheeler if you are interested in statistical process control. His articles are available for free on the Quality Digest web site. They are short and easy to read. He stays true to Shewhart who ‘invented’ SPC and to Deming and others who understood and promoted Shewhart’s work.
A great start is Wheeler’s article entitled “Probability Models do not generate your data”
Probability Models Don’t Generate Your Data | Quality Digest
Thanks for these references.

#### riosimbolon

##### Starting to get Involved
but you can find stable graph but it is out of spec, the data plays around 4-8 mm but spec 10-15..

stability = normality

#### Mike S.

##### An Early 'Cover'
Trusted
I think earlier posts pretty clearly stated a stable process is not necessarily normally distributed.

Furthermore, neither stability or normality have anything to do with being within specification.