truthQ: truthQ

In benkeser/drinf: Computes doubly-robust estimators and doubly-robust confidence intervals

truthQ

Description

A function to generate an estimate of g0n and g1n that is equal to Q10 and Q20 plus an error term. The function is written for compatibility with the drinf.tmle function, i.e., as though it were a SuperLearner wrapper. If newX is two columns, returns Q1n, if four columns, returns Q2n.

Usage

truthQ(Y, X, newX, family, obsWeights, b0 = 1, ba = -0.25, b = 0.5,
  r = -1/4, cons = 2, ...)

Arguments

Y	The outcome
X	The predictors
newX	The prediction frame
family	Not used, just for compatibility
obsWeights	Not used, just for compatibility
ba	The effect of each treatment on the outcome. See makeData to see how it is used.
b	Interaction coefficient for L01 and L11
r	The exponent of the rate at which error goes to zero as function of n.

Value

A list with predictions equal to truth + error

benkeser/drinf documentation built on Oct. 22, 2023, 9:50 a.m.