Subscribe Bookmark RSS Feed

Stability test - question on estimated expiry

aparker01

Community Trekker

Joined:

May 13, 2015

Hello,

I'm having some trouble with the stability analysis in JMP.  I have two lots and when I use the Degradation platform to test for poolability it turns out they have different slopes and different intercepts.  JMP estimates the expiry to be at 38 months.  When I plot the lots individually with regression lines and 95% confidence intervals, I am getting an estimate of 32 months with the rate limiting lot.  I'm wondering if this is a common problem or if I am doing something wrong?  I would thing the regression lines and confidence intervals should be the same when JMP creates them individually for lots or I manually create them myself.

Any help would be greatly appreciated.

Thanks!

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

The answers do not agree because your two estimates come from different approaches. The Degradation platform uses a linear model that includes all of the data and, therefore, leads to smaller confidence interval estimates, which in turn leads to longer expiry. Your second approach separates the data by batch. The separate regression analyses based on reduced data sets leads to larger interval estimates and, therefore, shorter expiry.

Learn it once, use it forever!
2 REPLIES
Solution

The answers do not agree because your two estimates come from different approaches. The Degradation platform uses a linear model that includes all of the data and, therefore, leads to smaller confidence interval estimates, which in turn leads to longer expiry. Your second approach separates the data by batch. The separate regression analyses based on reduced data sets leads to larger interval estimates and, therefore, shorter expiry.

Learn it once, use it forever!
aparker01

Community Trekker

Joined:

May 13, 2015

Thanks for the reply Mark.  I've been playing with this all morning and I think I realized the issue.  The Degradation function determined that there were different slopes and intercepts between lots, so the lots were not poolable.  Degradation was estimating my expiry on the rate limiting lot using a one sided lower confidence interval and I was estimating the expiry on the rate limiting lot using the regression analysis with a 2-sided confidence interval.  If I decrease the alpha on the Degradation or increase it on the regression analysis, the expiries match exactly.

Thanks for your help!