There are two scenarios that I've encountered with a few different customers and I'm not sure how either one is handled.
First case is that a reliability test is conducted multiple times, each one can generate a model easy enough. But how can these be aggregated into a single model. The aggregated model should reflect a confidence interval based on the multiple different results.
The second case is that multiple reliability tests are conducted but conditions are changed, maybe by DOE. Here the question is, which results are statistically different from the others? If the best result is statistically different, is there a confidence interval that describes how much better?
