Conditional quantile by inverting a kernel CDF

\(\hat q_\tau(x) = \inf\{y: \hat F(y|x) \ge \tau\}\). All tau come from ONE conditional CDF, so the estimate is automatically monotone in tau – the crossing that afflicts independently fitted quantiles cannot arise. Mirrors morie.fn.hrzkqre.

Usage

morie_kernel_quantile(x, y, tau = 0.5, grid = NULL, h = NULL)

Arguments

x, y

numeric regressor and response.

tau

quantile level(s) in (0, 1).

grid

evaluation points.

h

bandwidth.

Value

list: grid, quantile, tau, bandwidth, monotone_in_tau.

References

Horowitz, Ch. 3.

Examples

set.seed(1)
morie_kernel_quantile(runif(200), rnorm(200), tau = c(0.1, 0.9))$monotone_in_tau
#> [1] TRUE