ivqr_confint: Compute the confidence interval for Betaj

In ChenyueLiu/ivqr: Perform inverse quantile regression by solving mixed integer linear programming

Description

This function is an internal function called inside function "ivqr_fit()"

Usage

ivqr_confint(test_stat = test_stat_freq_D, j = 1, B = NULL,
  Phi = NULL, Beta_j_hat = 1, Srate = 1/2, alpha = 0.1, Y, D, X, Z,
  tau = 0.5, width_ratio = 1, M = 10, homoskedastic = FALSE,
  kernel = "Powell", lpsolver = NULL)

Arguments

D	data for endogenous variable D
X	data for exogenous variable X
Z	data for instrumental variable Z
tau	tau in quantile regression
M	a large number
lpsolver	"gurobi","cplexapi","lpsolveapi"

Value

the estimates, upper bound and lower bound

Examples

n=50
pD=3
sample_data<-chen_lee(n,pD)
Y=sample_data$Y
D=sample_data$D
Z=sample_data$Z
X = matrix(1, n, 1)
XZ = cbind(X, Z)
PI_XZ = XZ %*% solve(t(XZ) %*% XZ) %*% t(XZ)
Phi_CH = PI_XZ %*% D
lpsolver="gurobi"
tau = 0.5
iqr_milp_fit = iqr_milp(Y, D, X, Z, tau = tau, lpsolver = lpsolver)
B_D_hat = iqr_milp_fit$B_D
J=1
ivqr_confint(
test_stat = test_stat_freq_D,
j = J,
B = Phi_CH[, J, drop = FALSE],
Phi = Phi_CH[, -J, drop = FALSE],
Beta_j_hat = B_D_hat[J],
Srate = 1 / 2,
alpha = 0.1,
Y=Y,
D=D,
X=X,
Z=Z,
tau = tau,
M = 10,
homoskedastic = FALSE,
kernel = "Gaussian",
lpsolver = lpsolver
)

ChenyueLiu/ivqr documentation built on Aug. 9, 2020, 7:49 p.m.