simpleDataB: Simple simulated data: Effect modification and no confounding

Description Usage Format Estimated subgroup assignment References See Also

Description

A simulated dataset of 1500 observations, with 'observed' binary treatment, Normal outcome, covariates, and a few other variables that were used in the generation of the 'observed' treatment and outcome.

Usage

1

Format

A data frame with 1500 rows and 16 variables:

trt

treatment, generated from Bern(expit(0.4 X5))

Y

outcome, corresponding to the potential outcome (Y0 or Y1) associated with the observed treatment

X1

confounder, generated from N(0,1)

X2

confounder, generated from N(0,1)

X3

confounder, generated from N(0,1)

X4

confounder, generated from N(0,1)

X5

instrument, generated from N(0,1)

X6

prognostic variable, generated from N(0,1)

E1

effect modifier, generated from Bern(0.5)

E2

effect modifier, generated from Bern(0.5)

E3

effect modifier, generated from Bern(0.5)

Y0

potential outcome, generated from N(-3.85 + 5 trt + X6 - E1 - 2E3 + trt*E1 + 4 trt * E2 - 4 trt*E3, 1 ) where trt is set to 0

Y1

potential outcome, generated from N(-3.85 + 5 trt + X6 - E1 - 2E3 + trt*E1 + 4 trt * E2 - 4 trt*E3, 1 ) where trt is set to 1

trueGrp

the true average treatment effect in the subgroup that the observation belongs to

estGrp

an integer value in {1, 2, ..., 10} that represents the estimated subgroup that the observation belongs to. See below for more information on how this value is assigned.

mmt

the estimated ITE, as determined via regression using Bayesian Additive Regression Trees (as estimated by bart() of the bayesTree package).

Estimated subgroup assignment

To estimate what subgroup each observation belongs to, an individual treatment effect (ITE) was estimated (stored in mmt). This empirical distribution of ITEs is then partitioned into deciles, and an average treatment effect (ATE) estimated in each. Each of the ten subgroups are arranged in ascending order by the subgroup-specific ATE then assigned an integer value in {1, 2, ..., 10} (stored in estGrp).

References

Anoke SC, Normand S-L, Zigler CM (2017). Approaches to treatment effect heterogeneity in the presence of confounding (in revision).

See Also

simpleDataA for data simulated with confounding and no effect modification, simpleDataC for effect modification and confounding, and simpleDataD for effect modification and confounding by effect modifiers.

Other simple simulated datasets: simpleDataA, simpleDataC, simpleDataD


sanoke/hetviz documentation built on March 4, 2020, 7:58 a.m.