RIOM: Function for calculating warning accuracy and sensitivity...

In siyuheng/RIOM: Randomization Inference with Outcome Misclassification

Function for calculating warning accuracy and sensitivity weights in randomization inference with outcome misclassification

Description

Function for calculating warning accuracy and sensitivity weights in randomization inference with outcome misclassification, which works for both Fisher's sharp null and Neyman's weak null.

Usage

RIOM(treat.ind, outcome.ind, index, alpha = 0.05, null.hypothesis, type.random, timelimit = 1000, gap = 0.00001)

Arguments

treat.ind	A N-length vector of the treatment indicators of all N units in the study, 1 if treated and 0 if not. The n-th entry is the treatment indicator of unit n.
outcome.ind	A N-length vector of the measured binary outcomes. The n-th entry is the measured outcome of unit n.
index	A N-length vector: the n-th entry is the index of the stratum of unit n. Suppose that there are I strata in total, then the value of each entry is an integer between 1 and I.
alpha	The alpha level of the two-sided test. The default is 0.05.
null.hypothesis	The null hypothesis of interest. Two options: "sharp" = Fisher's sharp null; "weak" = Neyman's weak null.
type.random	The type of the randomization design. Two options: "1" = a type 1 randomization design; "2" = a type 2 randomization design. See Heng and Shaw (2022) for details.
timelimit	The limit for the runtime in seconds of the function. The default is 1000 seconds.
gap	The tolerable gap of the upper bound and lower bound of the optimial solution. The default is 0.00001 (i.e., we want an exact warning accuracy).

Value

p-value	Two-sided p-value based on measured outcomes.
Difference-in-means Estimate	The point estimate of the average treatment effect based on measured outcomes (only for weak null).
Confidence Interval	The confidence interval of the average treatment effect based on measured outcomes (only for weak null).
Warning Accuracy	The warning accuracy given the observed data and alpha level.
Minimal Alteration Number	The minimal alteration number given the observed data and alpha level.
Sensitivity Weights	The four sensitivity weights given the observed data and alpha level.
Runtime	The total computation time in seconds.

siyuheng/RIOM documentation built on April 27, 2022, 12:05 a.m.