flipped.portkey: Flipped portkey two-coin algorithm: returns the result of the...

In dvats/portkey: Bernoulli Factory MCMC: Portkey Two-Coin Algorithm for Markov Chain Monte Carlo

Description

Flipped portkey two-coin algorithm: returns the result of the accept-reject step of a Flipped Portkey Barker's acceptance probability. Note: the flipped portkey is the same as Barker's for beta = 1.

Usage

flipped.portkey(prop, curr, beta = 0.99, pf, Cprop, Ccurr, ...)

Arguments

prop	The proposed value in the MCMC step
curr	The current value in the MCMC step
beta	The portkey parameter with default being .99. If beta = 1, this is the regular two-coin algorithm
pf	A function that produces a Bern(p) realization, with argument value for the state of the Markov chain
Cprop	Upper bound for the inverse target at the proposed value
Ccurr	Upper bound for the inverse target at the current value
...	additional arguments that go into pf

Value

a variable x which is either curr or prop and integer loops that returns the number of loops the Bernoulli factory took

Author(s)

Dootika Vats, [e-mail]

References

Vats, D., Gonçalves, F. B., Łatuszyński, K., Roberts G. O., Efficient Bernoulli factory MCMC for intractable posteriors (2020+), arxiv.

dvats/portkey documentation built on Oct. 16, 2020, 12:40 a.m.