Multi-way clustering

[MS4510]

I was wondering whether it's possible to use multi-way clustering when computing SE in pyblp, eg, cluster by market and product. I tried:

product_data['clustering_ids0'] = product_data['product_ids']
product_data['clustering_ids1'] = product_data['market_ids']

but got the following error "ValueError: The clustering_ids field of product_data must be one-dimensional." which led me to concldue this isn't possible, but wanted to double check. Thank you!

[jeffgortmaker]

Right, only one-way clustering is supported at the moment. You'd have to manually compute two-way clustered standard errors given the outputs of ProblemResults.

But this would be a great enhancement if you or someone else (or myself once I get some time) wants to add it. I think just having multiple columns of clustering IDs like that would be a really clean interface.

[MS4510]

Understood, thanks for confirming Jeff. Not sure I'm fit for the task but will keep it in mind ;-)