Bayesian Believe NetworksHas anyone done Bayesian believe networks in AgenaRisk? As far as I understand this is used in reliability and helps troubleshooting issues. It starts with mapping out the connections and probabilities for components that affect a product in nodes, then established connections between all the nodes (components) that are related. Engineers can answer a simple question for a node e.g. is this working or broken, and then the probabilities for other nodes are recalculated using a Bayesian approximation to show what the probabilities are that other parts are now going to pass or fail.
The reliability block diagrams look like maybe they do part of this? The interface for Agena looks more like a SEM diagram. However, the calculations and modeling in SEM is data heavy and confirmatory, whereas the BBNs are generative and can be used forward and backwards, and with very little (to no) data.
Has anyone experience with Agena, know if we do something like that in JMP, or if it would make sense to propose to do something like that in JMP if we are not?
Re: Bayesian Believe NetworksHi Monique. I have not used Agena, but I can tell you that Product Management has considered Bayesian Networks for possible future development. We've had some brief interaction with the development team about this, but no formal plans yet. Are you familiar with any customer requests for this functionality or have you heard any customers discuss this topic? Thanks, Sarah G
Re: Bayesian Believe NetworksSarah, yes, this was brought up by a customer! It seems this is widely used at their medical device organization (Medtronic) by engineers and scientists to gain a first step in root cause analysis. The customer who brought this up will be attending the Discovery Summit in Baltimore, if it would be helpful to gain more insight in how they use this at Medtronic. He has been a great advocate for JMP and JMP Live at Medtronic and thought this capability could be a great way to get more people to adopt JMP.
