Hypothesis Testing in DMAIC Projects

Six Sigma Methodology | Six Sigma Model | Six Sigma Approach
Knowledge Center

 

Next Topic

Six Sigma Methodology | Six Sigma Model | Six Sigma Approach > Best Practices > Hypothesis Testing in DMAIC Projects

Hypothesis Testing in DMAIC Projects
Zafar Kamal, Consultant, Canada, Member
Is hypothesis testing a must in DMAIC projects? Why (not)?
 

 
Hypothesis Testing in DMAIC?
Jens Folke, Consultant, Denmark, Member
Sometimes a DMAIC project can be based on Six Sigma process mapping only, i.e. SIPOC/IPO/C&E/FMEA, and the process check is a revisit to the RPN of the FMEA. In other cases you may not have a sufficiently large sample size to conclude, e.g. 2 proportions sometimes require thousands of samples, so the answer is no: not mandatory.
 

 
Hypothesis Testing in DMAIC
Zafar Kamal, Consultant, Canada, Member
Dear Folke, thanks for your comment, I agree 100%. What happens if we eliminate hypothesis evaluation for any data?
 

 
Yes - Hypothesis Testing is Needed
James, Business Consultant, United States, Member
If the solution is known, just implement it.
Otherwise DMAIC is a method to determine solutions when unknown. DMAIC uses various tools/techniques to sort out/drill down for potential causes. Then make a hypothesis ("Cause A impacts the result") and use facts/data to validate the cause before implementation. If you don't test your hypothesis the wrong potential solution may be implemented. HOW the hypothesis test is conducted may be a better question. But the idea behind hypothesis testing is valid for all DMAIC projects.
 

 
DMAIC is a Scientific Method
Manuel Razo, In-house consultant, Philippines, Member
DMAIC and scientific method are one and the same. So hypothesis testing is a must.
 

 
DMAIC is not only a Scientific Method
James, Business Consultant, United States, Member
DMAIC uses scientific methods, but involves more such as risk and change management.
 

 
Decision Making on an Incomplete Data Base
Jens Folke, Consultant, Denmark, Member
You have reduced the error rate from 5 in 180 til 3 in 200. A 2-proportion test is not significant, because you need about 3500 samples. This may take a year to get, so what do you do? Take a chance without a significant test, I presume. So hypothesis test was not mandatory in this case.
 

 
Depends on Cost and Benefits of Implementation
Jagdish B Acharya, Consultant, India, SIG Leader
Normal test of significance in hypothesis testing has 5% and 1% constants. In terms of deviation this amounts to two sigma and 2.5 sigma levels.
However if the cost of implementation (including risk) is less compared to expected value of benefits, the implementation may be done.
The design of test should also take care of interactions of factors and hidden factors which may have more contribution to variation.
 

 
Hypothesis Testing in DMAIC is not Always Necessary
Georges Van Cauwenbergh, Management Consultant, Ireland, Member
KISS... And by keeping it simple and having data for the total population (which by the way is not that unusual) we do not need statistics.
 

 
Data Always Leads to Statistics
James, Business Consultant, United States, Member
@Georges Van Cauwenbergh: Except when only provided one data point, we always use statistics to describe or understand the data. To know the average we calculate a statistic. We do not always choose or need to use inferential statistics, especially if we have the entire population. Not knowing your definition of population, I believe having the entire population is rather rare. Using data to make decisions means the data represents the future. The future data is part of the population and is unknown.
 

 
Samples and Populations and Hypothesis Tests
Ger de Waard, Management Consultant, Netherlands, SIG Leader
Why do we even need hypothesis tests? After all, we took a random sample and our sample mean of 325.8 is different from 250. That is different, right? Unfortunately, the picture is not so clear because we’re looking at a sample rather than the entire population. So we have "Sampling Errors" (SE).
The SE is the difference between a sample and the entire population. Thanks to SE, it's entirely possible that while our sample mean is 325.8, the population mean could still be 250. Or, to put it another way, if we repeated the experiment, it's possible that the second sample mean could be close to 250.

A hypothesis test helps assess the likelihood of this possibility and as we in LSS work with data and not always full populations I think using a hypothesis test is a requirement. But of course we don't do it by hand anymore we use tools like SPSS, Minitab, SigmaXL or QImacros to do the job for us and as a previous commenter stated KISS is always a good advisor.
 

     
Special Interest Group Leader
Jagdish B Acharya
Consultant

Six Sigma Methodology | Six Sigma Model | Six Sigma Approach
Summary
Forum
Best Practices


Six Sigma Methodology | Six Sigma Model | Six Sigma Approach
Knowledge Center

 

Next Topic



About 12manage | Advertising | Link to us / Cite us | Privacy | Suggestions | Terms of Service
© 2019 12manage - The Executive Fast Track. V15.1 - Last updated: 26-8-2019. All names ™ of their owners.