What are the Best Variable Data Statistical Tools to Use in Validating a Manufacturing Process?
I've been fortunate to have worked in many different industries for almost 40 years, so I've seen a wide variety of Process Validation approaches using different statistical tools.  Since 1996, I've served in leadership and consulting roles in Medical Device for Medtronic and Stryker along with many small to large companies.  The list below is the present approach I use for validating manufacturing processes using Variable Data for comparisons such as run-to-run, lot-to-lot, different process steps, etc.. 

* Boxplots - A quick visual for comparing multiple datasets
* Anderson-Darling and Doornik-Hansen Test - To determine if a dataset can be treated as a Normal Distribution
* t-Tests (Normal) or 1-Sample Sign Test (Non-Normal) - To determine if the Mean or Median are significantly different
* Bartlett's (Normal) or Lavene's (Non-Normal) - To determine if the variances are significantly different
* Data Transformations - To convert Non-Normal Datasets into Normal Distributions.
* Histograms and Process Capability Tools - To determine if the long-term Process Capability Index demonstrates that a manufacturing process is Six Sigma Capable, High Capability, Low Capability or Not Capable.

What tools have you used for your company and industry?  

Thank you in advance for sharing your Process Validation expertise with me and the ASQ community.

Jim Steele
1 Replies
John Finley
10 Posts
That's a great list that I agree with whole-heartedly.  I also commonly used one and two ANOVA when comparing methods, additives or other proposed process improvements.  Going into detail on a complex scrap or rework problem using a combination of Pareto diagrams and p or np charts has been enormously useful.  Those tools helped us point the more powerful tools of variables control charts and even Design of Experiments in the right direction to do the most good.  

John Finley