Nathaniel

Not an economist, but I think this is a cool kind of model is under served in the python infrastructure. I've used the statsmodels version to interrogate questions about the...

Will do a bit of research later this week, and try to pick it up this weekend.

So I looked into this a little bit yday and I have a few questions re: (a) the implementation in the pymc labs post and (b) bayesian vars in the...

Looking at this Stan implementation of a VAR(1) hierarchicall bayesian model: https://rpubs.com/jimsavage/hierarchical_var We can see a panoply of priors and hyper priors of various dimensions. I can fit the model...

That's amazing work @jessegrabowski , you're 10 steps ahead of me on this! Those unit-circle plots in particular are amazing! I'm happy to bow out if you want to do...

I'd love some help on this and i'm glad you said it's "not obvious" because i was struggling a bit with it. I'm going to try again to work on...

That's a fair point. The lower data countries would be cool to try incorporate. I'll try play around with it again this weekend. Maybe aiming just to understand your code...

By the way, the function you use at.nn.conv2d has disappeared from the aesera docs.

I've added a draft pull request here to get the conversation started: https://github.com/pymc-devs/pymc-examples/pull/456 The basic functionality works to build a hierarchical model on the macro-economic data series, but it takes...

Just realised i hadn't shown an example in the notebook of a full hierarchical fit using the multivariate normal likelihood. It samples slowly and without many divergences but the convergence...