Section 4.4 algorithm, edge-kernel constant C_K = 3.4375, with the two documented errata followed per the paper's own Section 6.2 worked example (see the Python docstring of morie.fn.causrddh for the full derivation).
Value
List mirroring the Python payload: estimate,
h1, f_hat, sigma2, n_left_h1,
n_right_h1, mean_left_h1, mean_right_h1,
m3, h2_left, h2_right, m2_left,
m2_right, n2_left, n2_right,
n2_left_full, n2_right_full, r_left,
r_right, h_unregularized, kernel_constant,
n, method.
References
Imbens, G. and Kalyanaraman, K. (2009), NBER Working Paper 14726, Section 4.4 eqs (4.8)-(4.13) and Section 6.2; published as Review of Economic Studies 79(3), 933-959 (2012), doi:10.1093/restud/rdr043 .
Examples
set.seed(1)
x <- runif(200, -1, 1)
y <- 1 + 2 * x + 0.5 * (x >= 0) + rnorm(200, 0, 0.3)
h <- Causrddh(x, y, cutoff = 0)
is.numeric(h$estimate)
#> [1] TRUE