Overlay OOS datapoints in boxplot

[analicia]

Point markers are diamonds instead of crosses.

[twiecki]

Looking at this code I think it's time to refactor it. Specifically,

    df_weekly = timeseries.aggregate_returns(returns, 'weekly')
    df_monthly = timeseries.aggregate_returns(returns, 'monthly')

Should happen inside the plotting function and only returns be passed in. The slicing is a bit confusing and this refactor could remedy this.

[analicia]

@twiecki df_weekly and df_monthly are now only defined by the plotting functions that require them.

[twiecki]

LGTM, thanks!

[twiecki]

[twiecki]

@ahgnaw just noticed that we should probably add a legend as well explaining the red dots.

[analicia]

Unfortunately matplotlib adds the line through the marker in the legend. What do you think of the text? Alternative: "Out-of-sample returns"

[twiecki]

import matplotlib as mpl
mpl.rcParams['legend.handlelength'] = 0

Perhaps using that instead in a rc_context: http://matplotlib.org/api/matplotlib_configuration_api.html#matplotlib.rc_context

Also, the monthly color red is a bit hard to read with the red markers.

[twiecki]

I like out-of-sample data.

[analicia]

I'll try out a few different colors. The daily, weekly, and monthly markers could each be different colors.

[twiecki]

I think we'd rather switch the red color of the last bar.

[analicia]

here's yellow or green. I have a slight preference for the green, but either works.

[twiecki]

Sorry, I meant the actual fill color of the barplot, not the points.

[analicia]

Ah ok. I think this last. It's changed to a yellow that is consistent with the color palette. Thoughts?

[twiecki]

Perfect.