Calculates the augmented inverse probability weighted estimate (AIPW) of treatment success for each observation and treatment (i.e. on the level of a single unit). This method scales the estimated probabilities of success by the probability of being assigned the treatment, and weighted by a the conditional expectation of success from prior periods of an adaptive trial. The conditional expectation function used is a grouped mean by treatment arm.
get_iaipw(data, assignment_probs, periods, conditions, verbose)A tibble/data.frame, containing the data used in the Multi-Arm-Bandit, with new columns pertaining to the individual AIPW estimate for each person and condition, and probability of assignment for each treatment at each period.
A data.frame, data.table, or tibble containing input data from the trial. This should be the results of a traditional Randomized Controlled Trial (RCT). Any data.frames will be converted to tibbles internally.
A tibble/data.table containing the probabilities of being assigned each treatment at a given period.
Numeric value of length 1; number of total periods in the simulation.
Logical; whether or not to print intermediate messages. Default is FALSE.
The specification for the individual AIPW estimates can be found in Hadad et al. (2021). The formulas in equation 5, formed the basis for this function's calculations. Here the regression adjustment used is the grouped mean of success by treatment, up until the current period of estimation (so at period 5, the grouped mean would be calculated using the results from periods 1 through 4).
Hadad, Vitor, David A. Hirshberg, Ruohan Zhan, Stefan Wager, and Susan Athey. 2021. "Confidence Intervals for Policy Evaluation in Adaptive Experiments." Proceedings of the National Academy of Sciences of the United States of America 118 (15): e2014602118. tools:::Rd_expr_doi("10.1073/pnas.2014602118").