Use coords / xarray magic in dependent density regression

[MarcoGorelli]

xref #14

Is there a way to avoid having to do

beta = pm.Normal("beta", 0.0, 5.0, shape=(1, K))

?
If there was shape=K, I could put dims='K', but then

pm.math.dot(x, beta)

wouldn't work.

Any hints?

EDIT (Oriol): Addresses #99

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[OriolAbril]

You can use "one": [1] in coords dict, or any other name/coord value you may prefer, it just has to be there and have length one to act as the 1 in shape.

Reply via ReviewNB

[MarcoGorelli]

so simple, thanks!

[MarcoGorelli]

so much more readable with xarray magic...

Reply via ReviewNB

[OriolAbril]

:)

Updating the examples is being an amazing resource for me to both see the potential of xarray as well as where is more help and work needed. I don't need convincing on xarray usage, but there are still things where it surprises me.

As a side not, I think somehting like max_mixture_weights["K"].values will be equivalent tomax_mixture_weights.coords.to_index() . The distinction between coord and variable is still something I don't grok yet, especially because of non-indexing coords, but it is cool to be able to treat coords as variables in many cases

[fonnesbeck]

Thanks, @MarcoGorelli