Reliability Forecast

[#1]

I am trying to predict warranty returns but want to use the Bayesian estimation tools to update the warranty predictions. No matter what I do I can’t seem to get the warranty predictions to update after applying a Bayesian model. Is it possible to use the Bayesian tools with Reliability Forecast? If not, how would you recommend going about adjusting those return predictions for the updated posterior parameters?

[#2 · Accepted solution]

Thanks for bringing this to my attention!

Bayesian model shouldn't have been available in Reliability Forecast. It is not wired into the list of "Choose Distribution". I will put this for future consideration. So, it is not possible now for your first question.

I don't have a recommendation, but maybe some thoughts on a couple of scenarios.

First, if you have sufficient amount of data, i.e., return counts, non-Bayesian estimate should have given good predictions. If you have sufficient data, and also a strong prior, I suspect there will be a good chance the information conflicts. Then a question will be which to trust, data or prior? So I would prefer not using Bayesian model if there are sufficient return counts.

Second, if you don't have sufficient return counts, and you are interested in future return counts from what are still running. The last resort now is do-it-yourself by using simulation. That is what is done in the software as well. You need to check out one of the two resources, whichever is available to you:

1. Statistical Methods for Reliability Data, by Meeker and Escobar, 1998. Chapter 12.

2. Statistical Methods for Reliability Data, 2nd ed., by Meeker and Escobar and Pascual, 2022. Chapter 15.