get_bandit.ucb1

Calculates upper confidence bounds for each treatment arm

internal

Simulates the results of completed randomized controlled
trials, as if they had been conducted as adaptive Multi-Arm Bandit
(MAB) trials instead. Augmented inverse probability weighted
estimation (AIPW), outlined by Hadad et al. (2021)
<doi:10.1073/pnas.2014602118>, is used to robustly estimate the
probability of success for each treatment arm under the adaptive
design. Provides customization options to simulate perfect/imperfect
information, stationary/non-stationary bandits, blocked treatment
assignments, along with control augmentation, and other hybrid
strategies for assigning treatment arms. The methods used in
simulation were inspired by Offer-Westort et al. (2021)
<doi:10.1111/ajps.12597>.

Noah Ochital

whatifbandit

Analyzing Randomized Experiments as Multi-Arm Bandits

Ryan T. Moore

get_bandit.ucb1 function

<dl><dt>past_results</dt>
<dd>A tibble/data.table containing summary of prior periods, with
successes, number of observations, and success rates, which is created by <code>get_past_results()</code>.</dd>
<dt>current_period</dt>
<dd>Numeric value of length 1; current period of the adaptive trial simulation.</dd></dl>

Arguments

UCB1 Sampling Algorithm — get_bandit.ucb1

<dl>

<dt>past_results</dt>
<dd>A tibble/data.table containing summary of prior periods, with
successes, number of observations, and success rates, which is created by <code>get_past_results()</code>.</dd>


<dt>current_period</dt>
<dd>Numeric value of length 1; current period of the adaptive trial simulation.</dd>

</dl>

get_bandit.ucb1: UCB1 Sampling Algorithm

Description

Usage

Value

Arguments