Truncated Data in Reliability AnalysisHey all! I got a really interesting question today:
" I am working on a reliability analysis where the outcomes are NOT limited to Fail or Survival.
There are three options:
- FAIL
- SURVIVE (censored usually at a given time specification, 1000 hours, etc.)
- QUIT (where the unit or person simply leaves the process without incurring a FAIL)
I am impressed with the scope of the Reliability Modules available. Can you help me expand their application into this problem definition?"
Mark reminded me that that third option is a truncated data point, but I can't find any references on how to treat that data with JMP. Thoughts from the group?
Best,
M
Re: Truncated Data in Reliability AnalysisTobias & Trindade (Applied Reliability, 3rd Ed) have a discussion of modeling truncated data (Section 4.5 and 9.8). They use Excel and say it can be done in JMP although they don't provide details.
I agree with Jerry, I think you're dealing with censoring and not truncation. From Tobias & Trindade (p43):
"The important difference between censored and truncated data is that truncated items (typically an unknown number) are excluded from the sample under observation. In contrast, censored observations are included in the study as part of the sample."
Re: Truncated Data in Reliability AnalysisCensoring can occur at any time. Doesn't have to be a constant value for all parts under test (for example). e.g. if testing motors, you may plan to run the test for 5000 hrs (so any motor that survives to 5Khrs is considered censored.) But if a motor driver (not the motor itself) failed after 1000 hrs, and it caused you to remove the motor from the test, you would just list that motor as censored after 1000 hrs.
The motor survived 1000 hrs, so you don't want to lose that data, so it's perfectly fine to include it as censored at 1k. JMP can handle it!
