Ordinary-smooth error gives \(O_p(n^{-r})\); supersmooth error
gives \(O_p[(\log n)^{-s}]\). The gap is the practical message of
the chapter: at \(n = 10^6\) a logarithmic rate has barely moved.
Both are returned at the requested n so the difference is a number
rather than a footnote. Mirrors morie.fn.hrzdcrc.
Usage
morie_deconv_rate(n, error = "normal", s = 2, r = 2)
Arguments
- n
sample size.
- error
"normal" or "laplace".
- s
supersmooth exponent.
- r
ordinary-smooth exponent.
Value
list: rate, regime, polynomial_rate, logarithmic_rate,
ratio, n, method.
References
Horowitz, Ch. 5, Sec. 5.1.1-5.1.2.
Examples
morie_deconv_rate(1e6)$ratio
#> [1] 5239213806
