search

quantopian / bayesalpha

Bayesian models to compute performance and uncertainty of returns and alpha.
Other
106 stars 56 forks source link

MAINT: author model should not rely on hard-coded column names #55

Closed eigenfoo closed 6 years ago

eigenfoo commented 6 years ago

Currently, the author model takes a DataFrame that must have the column names meta_user_id, meta_algorithm_id, meta_code_id and perf_sharpe_ratio_is. While the first three column names might be reasonable (since they're IDs), the last one certainly isn't. What if we want to feed in different sharpes (e.g. specific, oos, etc.)?

It should be able to pass in a parameter called data_column_name. Otherwise, we fall back on the perf_sharpe_ratio.

twiecki commented 6 years ago

I think just requiring calling it sharpe_ratio is reasonable.

eigenfoo commented 6 years ago

Closed by #58