Re: Purpose of using different Distributions?
Hi,
I do not understand why sometimes we choose weibull and sometimes binomial and sometimes other distributions to solve reliability related problems?
WHat is the reason of first choosing a distribution and then solving for the reliability?
There are two primary reasons for fitting a distribution:
1. I have a theory related to my engineering and operational knowledge of the system being measured. For example, do I expect the failures to be "memory-less" (exponential) or do I expect there to be a "burn-in" period and/or a "burn-out" period (the "bathtub curve", generally fit with weibull).
2. Analysis of the empirical data shows me that one curve fits better than the other.
What distribution that is fit is especially important with high-reliability, low failure rate systems. I am many times trying to estimate the reliability of the system out many standard deviations, yet only have a limited set of data, which may be truncated data (not running the system sufficiently long enough to get an actual failure).
