Fixing cases when ValueError from alphalens.tears.create_factor_tear_sheet

[dat-boris]

When we run create_factor_tear_sheet on either:

1. single value for factor per group (only one value over a date, or sector)
2. Field with discreet value (not sure if this is a correct use case for alphalens)

We are can run into exception seen below.

qcut over similar edges have been discussed but no solution is implemented yet "officially":

• http://stackoverflow.com/questions/20158597/how-to-qcut-with-non-unique-bin-edges
• qcut() should make sure the bins bounderies are unique before passing them to _bins_to_cuts pandas-dev/pandas#7751 (comment)

So this provided a simple work around for case when we run into this issue in a simple case, when we have only seen one value of the factor.

Exception seen

/usr/local/lib/python2.7/dist-packages/alphalens/performance.pyc in quantile_calc(x, quantiles)
    242 
    243     def quantile_calc(x, quantiles):
--> 244         return pd.qcut(x, quantiles, labels=False) + 1
    245 
    246     grouper = ['date', 'sector'] if by_sector else ['date']

/usr/local/lib/python2.7/dist-packages/pandas/tools/tile.pyc in qcut(x, q, labels, retbins, precision)
    167     bins = algos.quantile(x, quantiles)
    168     return _bins_to_cuts(x, bins, labels=labels, retbins=retbins,precision=precision,
--> 169                          include_lowest=True)
    170 
    171 

/usr/local/lib/python2.7/dist-packages/pandas/tools/tile.pyc in _bins_to_cuts(x, bins, right, labels, retbins, precision, name, include_lowest)
    187 
    188     if len(algos.unique(bins)) < len(bins):
--> 189         raise ValueError('Bin edges must be unique: %s' % repr(bins))
    190 
    191     if include_lowest:

ValueError: Bin edges must be unique: array([-0.0030303, -0.0030303, -0.0030303, -0.0030303, -0.0030303,
       -0.0030303])

[jameschristopher]

So we have ran into this qcut issue as well, the first and most obvious solution is to reduce the number of quantiles, but it looks like in from your trace (and description) that you are assigning groups of size N >> 1 a value and trying to look at the performance of each group? So are you classifying groups of stocks into varying degrees of under/over performing?

[twiecki]

Closing due to inactivity.