factor_agent.config.AdmissionConfig#

class AdmissionConfig(t_threshold=3.0, max_correlation=0.7, min_monotonicity=0.6, min_coverage=0.5, trial_exponent=1.0, confirm_t=2.0)[source]#

Bases: object

What an evaluated candidate needs to be accepted.

Attributes#

AdmissionConfig.confirm_t = 2.0#: Newey-West t of the partial IC, in the factor’s direction, a candidate needs on the confirmation segment (segments.confirmation), whose statistics no prompt shows; read only when there is one.

AdmissionConfig.max_correlation = 0.7#: Largest absolute cross-sectional correlation with a reference or accepted factor.

AdmissionConfig.min_coverage = 0.5#: Smallest mean daily share of the members on which the candidate has a value.

AdmissionConfig.min_monotonicity = 0.6#: Smallest Spearman correlation of the decile returns with the decile number.

AdmissionConfig.t_threshold = 3.0#: Newey-West t of the partial IC the first evaluated trial needs; the threshold rises with the trial count (Bonferroni, see _admission).

AdmissionConfig.trial_exponent = 1.0#: Power of the trial count the anchor’s p-value is divided by: 1 is Bonferroni, a smaller one lets the threshold rise more slowly (0.5 counts 100 trials as 10).