cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
0 Kudos

Sequential A/B Testing

What inspired this wish list request? Conventional fixed horizon A/B testing such as a t-test for two independent populations or a z-test for a a population against a known mean are powerful but can take a long time to complete.  Considering the z-test, when the initial data points of the population under test clearly show a difference compared to the reference mean, the full pre-defined sample size may not be needed to prove a difference in the origin of the populations. Here, the sequential A/B testing methodology can speed up the expeirment by ending it early yet still have the statistical rigor required.  Additionally, the dynamic sequential testing capability exists in other statistical software tools yet is not available in JMP.  As a JMP user, I would like this capability.

 

 

What is the improvement you would like to see? I would like the JMP team to build a new feature for A/B sequential testing as part of their DOE platform. I would like to see a statistical algorithm built that analyzes data sets against the efficacy and futilitiy boundary as data poit.  References to the literature can be found in the wikipedia page: https://en.wikipedia.org/wiki/Sequential_analysis

 

 

Why is this idea important?  This feature would add to the power of JMP for process engineers and scientists who design experimental testing for new process change improvements.  The ability to statistically decide to end an experiment early will benefit the JMP user community.     

 

 

1 Comment
SarahGilyard
Staff
Status changed to: Acknowledged

Thank you for this suggestion. We are discussing this internally and will reach out when we know more. (PM and sales met with this customer on March 19th to discuss).