quantlab.model.evaluation.information_ratio#

quantlab.model.evaluation.information_ratio(series)[source]#

Return the mean of the finite values of series over their sample std.

This is the ICIR when series is a per-bar IC series. NaN values (the skipped bars of an IC series) are left out rather than counted as 0. The standard deviation uses ddof=1, so fewer than two finite values give NaN, and so do finite values that are all equal (tested exactly, as for the IC’s constancy test), since the ratio is then undefined.

Parameters:

seriesarray_like: A 1-D sequence of per-bar values.

Examples

>>> information_ratio([0.1, np.nan, 0.3])
1.4142135623730951
>>> information_ratio([0.1, np.nan])
nan