BUG? plot_quantile_returns_violin and than plot_quantile_returns_bar use different data about alphalens HOT 3 CLOSED

As a side note I would probably drop the calculation of mean_ret_quant_daily, it doesn't suit well with all possible data frequency passed to alphalens. Or, at lest, we should avoid mixing plots where the data is aggregated by day (e.g. cumulative plot by quantile, top minus bottom quantile mean return) with plots where it is not (e.g. factor weighted cumulative plot). This is confusing.

from alphalens.

To summarize, I believe the aggregation of data by day is a nice feature that would produce better plots for some scenarios (data frequency/data time span) but not for all the possible input of alphalens. This means the aggregation feature should be optional and the aggregation period (day, hour, week) configurable. Also, to be useful, the feature should be applied to all the plots consistently.

Currently the aggregation of data by time rule is not optional, it is not configurable and it is applied randomly to some plots only. This is confusing and I would completely remove the feature until properly implemented.

from alphalens.

The code doesn't exactly does what I thought and it's almost fine. Still it doesn't make sense to aggregate by day when calculating perf.mean_return_by_quantile for the reasons exaplained above.

from alphalens.