get_eps_zero: Estimation of the minimum proportion of confounded units such...

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

Description

get_eps_zero computes an estimate of eps0 and Wald-type confidence interval.

Usage

get_eps_zero(
  n,
  eps,
  pt_est,
  lb,
  ub,
  ql,
  qu,
  ifvals_lb,
  ifvals_ub,
  delta = 1,
  alpha = 0.05
)

Arguments

n	sample size.
eps	vector of arbitrary length specifying the values for the proportion of confounding where the lower and upper bounds curves are evaluated.
pt_est	point-estimate under no unmeasured confounding. Generally, mean(nuhat), where nuhat is a vector of influence fn values for the parameter EE(Y|A = 1, X) - E(Y|A = 0, X).
lb	length(eps) x (num of delta values) matrix containing the values for the lower bounds.
ub	length(eps) x (num of delta values) matrix containing the values for the upper bounds.
ql	a list of size nsplits, where element j is a (num eps) x (num delta) matrix containing estimates of eps-quantile of g(etab) for lower bound computed using all obs except those in fold j.
qu	a list of size nsplits, where element j is a (num eps) x (num delta) matrix containing estimates of eps-quantile of g(etab) for upper bound computed using all obs except those in fold j.
ifvals_lb	a list of size nsplits, where element j is n x length(eps) x length(delta) array containing values for ifvals_lb - lambda_lb * q_lb computed using regression functions estimated from all obs except those in fold j and evaluated at obs in fold j.
ifvals_ub	a list of size nsplits, where element j is n x length(eps) x length(delta) array containing values for ifvals_ub - lambda_ub * q_ub computed using regression functions estimated from all obs except those in fold j and evaluated at obs in fold j.
delta	vector of delta values specifying the values of delta used to bound maximal confounding among S = 0 units. Length(delta) should match the 2nd dimension of lb, ub, ql, qu and the 3rd dimension of ifvals_lb and ifvals_ub.
alpha	scalar specifying the confidence level. Default is 0.05.

Value

A length(delta)x5 data.frame with values of delta, estimate of eps0, max(0, ci_lo), min(1, ci_hi), variance of estimate of eps0.

See Also

if_gamma, if_tau, get_ifvals.

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