Problem with Lower Limits of the Stability Predictor triggering at T0 in JMP PRO

[#1]

Hi All,

Hope we're all ok!

So, when using the Stability test (in the Degradation Data Analysis module of JMP PRO, 16.0.0.0)  I've come across something I'd like help/opinions on. I'm finding that when I have a particularly variable data set the CI range pass the lower limit at the first data point (i.e. at T0). Example screen-grab attached.

Now I know for this example data set that it is impossible for these data to go below zero. So to get the prediction I'm after I've been artificially lowering the lower-limit so that it doesn't trigger at T0. 

My question is this; is there a better way to do this?

Can I 'tell' the model that going below one of the levels is impossible, and as such not to 'allow' it? I obviously don't want the module to always do this as I also have use cases where crossing either level is valid (and I'd need to know), but in some cases (such as this one) I know that one level would never be crossed.

Cheers,

Jon

[#2]

Hi @Jon_Armer ,

  Since you know that you have a data set that can never have values less than zero, you might want to consider performing a transformation, e.g. log, or something like that. This way, when you perform the analysis JMP will be able to account for the positive only values for your fit and the model will represent something closer to what you'd expect. You can read a nice presentation here on it. There was also a similar post here about it. There's also a good SAS blog about it here.

Hope this helps!,

DS

[#3]

I think that you need to remove the lower limit completely. I.e. put a dot in lower limit for missing value.
In addition, choice between using one limit and two limits has an impact on your analysis. For example, your analysis should only report the earliest crossing time on the upper limit as an expiration day, but you nevertheless put a lower limit in the analysis, the resulting number is not what you want.
So you need to deliberately decide to enter one limit or two, and they shouldn't be manipulated arbitrarily.
See the third paragraph on this page to see how one limit and two limits determine what values are calculated: https://www.jmp.com/support/help/en/16.2/#page/jmp/stability-analysis.shtml#