numeraire.AverageAbsAlphaEvaluator#

class numeraire.AverageAbsAlphaEvaluator[source]#

Bases: object

Average absolute pricing error (mean over assets of |mean realized - mean predicted|).

Each asset’s alpha is its mean realized return minus its mean predicted expected return; the metric is the cross-sectional mean of the absolute alphas (in the input’s return units). The magnitude companion to CrossSectionalR2Evaluator. (Factor-model joint zero-alpha inference stays in numeraire.core.stats.grs_test(), which needs the factor returns this generic pricing surface deliberately does not assume.)

__init__()#

Methods

__init__()

evaluate(oos_output)

Attributes

requires