Comparing Life Data Sets

[#1]

Hello, Some statistical SW packages offer the ability to compare two or more sets of reliability or life data in order to determine which of the data sets has a more favorable life distribution. The data sets could be from two alternate designs, manufacturers, lots, assembly lines, etc. The data sets may contain censored data. In general, the problem boils down to that of being able to determine any statistically significant difference between the two (or more) samples of potentially censored data from two possibly different populations.

These tools comes to mind:

(a) Contour Plots: To determine whether two data sets are significantly different and at what confidence level, one can utilize the contour plots. By overlaying two contour plots from two different data sets at the same confidence level, one can visually assess whether the data sets are significantly different at that confidence level if there is no overlap on the contours (were the same distribution must be fitted to both data sets).

(b) Life Comparison Tool: suggested by Gerald G. Brown and Herbert C. Rutemiller, to estimate the probability of whether the times-to-failure of one population are better or worse than the times-to-failure of the second (could be from differnt distributions). The same tool could be used to perform Stress-Strength analysis.

Does JMP13 (or future version 14) provide any of these tools in their standard or Pro versions?

Zvika Bar-Kochva

Philips Medical Haifa

Israel

[#2]

Hi Zvika

Possibly the Life distribution platform is what you are looking for.

Have a look at: https://www.jmp.com/support/help/13-2/Example_of_the_Life_Distribution_Platform.shtml

Kind Regards, Thomas

[#3]

Hi Tomas,
I searched JMP's Reliability and Survival book and went through the suggested demonstration files including the one you mentioned.Unfortunately, I found no obvious built in solution.

Regards, Zvika

[#4]

Hi @Zvika_BK,

I'm pretty confident this type of contour plot is not a built-in functionality, but with a little elbow grease you could probably write the functions for the life distribution densities as functions of their parameters and plot them in Contour Plot.

Comparing the density functions is pretty easy using Fit Life by X and Fit Parametric Survival.  Both have profilers of the distribution function so you can toggle between the factor levels and see the impact on the distribution.

[#5]

The contour plots for the parameter estimates are available through the red triangle menu of the Parametric Estimate report:

The result is:

The comparison does not directly overlay the contours for the estimates of different groups, though.

[#6]

Hi Mark, Thanks for your answer. I was aware JMP can show likelihood contour charts for individual distributions. However, I was looking for a (quick and simple) method to overlay two or more such plots on a single chart, because (a) It strikes me as a natural capability of any statistical SW package designed for reliability engineering, especially one intended and designed for providing means for comparing reliability distributions. (b) I’ve found this representation method very useful and illuminating, especially when one need to communicate results to management or other engineers (generally people with less solid statistical background). (c) Most statistical SW packages, including JMP, will provide many methods of comparing the mean of distributions, however, (modern) reliability design practices, comparing the mean is not very useful since we are interested much more in the tails of distributions or in the effect of the entire distribution. Zvika

[#7]

It is possible to overlay the two contours, but unfortunately I couldn't find a way to change colors and transparencey to make it look nice... 

Right click one of the countours and choose 

edit->Copy frame contents

then on the other:

edit->paste frame contents.

And make sure both are within the axis scales.

[#8]

OK. I see your point. Thanks.
I guess it's better than nothing at all but really far from satisfactory.

The list of grievances below are intended to JMP developers, not you:
(a) One cannot control the number of confidence level plots shown (even for a single distribution). It always present 6 plots for confidence levels in the range 0.5 to 0.99.
(b) Some plots are cut when reaching the end of the auto-selected scale. However, these are not re-drawn once the scales are manually adjusted.
(c) As you said, there is no way to control colors and identification of the plotted curves.
(d) It’s a drag to do it over and over again if you are comparing several distributions and/or performing multiple analyses.

[#9]

Your points are well-taken. You can currently compare the likelihood contours of different groups with side by side plots, not overlaid plots.

Please add your suggestion to our Wish List. You will find a button in the menu at the top of this page.

[#10]

Is there an update to the contour plots where different distributions can be compared?