Hi,
a customer pointed me to something I believe are different calculation methods. But I want to be sure and therefore ask more experience people with life distributions 
In Minitab the comparison of Weibull and TH Weibull is showing a threshold parameter of about 10.
In JMP for the attached table the Weibull and TH Weibull is almost the same and the threshold parameter close to zero. 
If I exclude values with Time <9 as done in the attached table using Column 4 indicator, the threshold parameter is about 9 close to what Minitab is doing with the whole dataset. From the documentation in both Minitab and JMP I can see that Minitab says it is estimating the threshold where the first failure may occur, in JMP it takes the first failure and calculates alpha and beta for the rest. Is that true? So the threshold in JMP is just the first actual occuring failure and not the estimation when it may occur looking at the whole data of failures? This would at least explain the missing difference (against expectations) in Weibull and TH Weibull in this case as there are some low failure times.
(If I exclude with Column 4 values below 9 I get a threshold parameter of about 9,5 and it differs from Weibull distribution but couldn't converge, but the fitted distribution looks much closer to what one would probably expect)
If that's the case, is this the standard approach or do we do something that is unique to JMP or just another approach?
Thanks for your input!
Martin
