if_tau: Estimation of influence function for E(g(eta))

In matteobonvini/sensitivitypuc: Sensitivity Analysis via the Proportion of Unmeasured Confounding (PUC)

Description

if_tau returns the influence function values for the parameter E(g_l(eta)) (upper=FALSE) or E(g_u(eta)) (upper=TRUE) at each X.

Usage

if_tau(data, delta, upper = FALSE)

Arguments

data	data.frame of dimensions n x 8, with columns: y, a, ymin, ymax, pi0g (= P(A = 0 | X = x) if model "x" or 1-A if model "xa"), pi1g (= P(A = 1 | X = x) if model "x" or A if model "xa") mu1 (E(Y | A = 1, X = x)) and mu0 (E(Y | A = 0, X = x)).
delta	vector of values for the delta parameter.
upper	boolean (default is upper=FALSE) if TRUE if values for E(g_u(eta)) are returned, otherwise those for E(g_l(eta)) are returned

Value

a length(y)xlength(delta) matrix containing the influence function values at each observed X of the parameter E(g(eta)).

Details

As done in the paper, one can see that g(eta) for the lower bound is equal to g(eta) for the upper bound minus delta * (ymax - ymin). Therefore the IFs for E(g(eta)) follows the same relation. They are keep separated just for code clarity.

See Also

if_gamma, get_ifvals.

matteobonvini/sensitivitypuc documentation built on Dec. 9, 2020, 2:24 a.m.