adastop.Bounded_Confseq

class adastop.Bounded_Confseq(delta=0.05, bounds=(0, 1), c=0.75)

Bases: object

Compute a confidence sequence for bounded data using the hedged betting confidence sequence from Section 4.4 of Estimating means of bounded random variables by betting. JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B by Waudby-Smith and Ramdas (2022)

Parameters:

delta: float, default=0.05

Type I error. The resulting confidence interval will be of level 1-delta

bounds: tuple of floats, default=(0,1)

Bounds on the support of the distribution of values.

c: float, default=3/4

Tuning parameter for the betting process, measure how conservative the confidence sequence is. 3/4 is a good default.

Attributes:

self.stats: dictionary with keys “est”: the estimation, “lb”: lower bound of confidence sequence, “ub”: upper bound of the confidence sequence and “N”: number of samples.

Methods

capital
observe

__init__(delta=0.05, bounds=(0, 1), c=0.75)

Methods

__init__([delta, bounds, c])
capital(m)
observe(x)