# Working with Non Normal data

Six Sigma – iSixSigma Forums Old Forums General Working with Non Normal data

Viewing 8 posts - 1 through 8 (of 8 total)
• Author
Posts
• #47448

Participant

Hi,Confused about how to prcoeed when working with tensile data for a component, this is typicaly non normal and approximates a 3 paraamter weibull better than a normal distribution. Given that situation I am unsure if Minitab can offer me full use of capabiliy anyalysis, and control charting as this is tradtionally premised on normal data.

0
#158190

Ashman
Member

Normal data is the exception.  Use Shewhart Charts … they do not rely on normality.

0
#158194

Jim Shelor
Participant

John,
Minitab’s capability analysis has a non-normal selection and 3 parameter weibull is one of 13 distributions you can use for this analysis.
The distribution ID function in Minitab will tell you what your distribution is.
The control charts in Minitab have a selection under options to perform a Box-Cox transformation on the data when performing the control chart.
I hope this helps.
Jim Shelor

0
#158195

Jim Shelor
Participant

Steve,
Shewhart’s control charts for variable data do rely on normality.  The reason is all the constants used to determine control limits and other parameters are based on normal data.
You can plot non-normal data, but the answers you get from the control chart are not going to be reliable.
Regards,
Jim Shelor

0
#158203

accrington
Participant

Not this old chestnut again? Chapter 4.3 in Wheeler and Chambers Understanding Statistical Process Control, gives specific examples of  how robust the charts are to non – normality. Also, remeber the Central Limit Theorem!

0
#158384

Jim Shelor
Participant

Accrington,
Aside from the X-bar chart, where does the Central Limit Theorm apply in control charting?
Jim Shelor

0
#158386

Participant

John
You say the data is “typically” non normal. What does that mean? Is the process so out of control that the data appears to be non normal? I agree with Jim S. Run the Box-Cox Transformation and see what the data tells you from there.

0
#158463

mcleod
Member

John;
The first thing I would stress is to understand why the data is non-normal. Bi-modal distributions come from different populations many times- therefore combining them may not give you a true understanding of what is going on in your process. There are no transformations to handle bi-modal distributions unless they are very close to normal to begin with.
If your data is skewed and not bi-modal, you may also try the Johnson transformation. Minitab will allow a capability analysis using the Johnson transformed data as well – however, you need to make sure that your limits are real after the transformation.
Hope this helps.
Scott

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

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