Closed hyunjimoon closed 3 years ago
Thanks @hyunjimoon! Have you received any feedback on this case study yet? If not I can try to find someone to give you feedback on this who is more of a gaussian process person than I am.
I have received the following feedback through the mail from Ben Bales, Michael Betancourt, Bob Carpenter. There were some conflicting points but did my best to address them.
keep them simple and focused, could use simulated data as engine failure details are more distracting than helpful -> Made amends by deleting some models. I tried to focus on following the Bayesian workflow.
add data description and goals of analysis -> added link to stancon talk youtube and inserted the following:
This casestudy shows how identification and underfitting problems diagnosed from pushforward and predictive checks are addressed through reparameterization and adding variables. Basically our data is highly unbalanced per category with lots of missing data. Also, due to its hierarchical structure of a system, such as shared engine types, hierarchical model is applicable.
I like case studies with real data. If the engine failure data is a nice illustration of some aspect of GPs, that's great---it doesn't have to be the best possible model for the engine failure data.
Ok great, just wanted to make sure that at least one person gave you feedback. Will merge this now!
add gaussian process case study description