raw
fun rawMomentInLogit(j: Int, endpointTolerance: Double = DEFAULT_ENDPOINT_TOLERANCE, quantileOfLogit: (Double) -> Double): Double(source)
The j-th moment about the origin, for a quantile function parameterized by the logit of the cumulative probability rather than by the probability.
Prefer this over the probability-based overload whenever the far tail carries weight. A probability supplied as a double cannot recover its own logit beyond about ten to the minus tenth from either endpoint, which makes the integrand noisy in the tail and can prevent the quadrature from converging at all.
Parameters
j
the order of the moment, which must be at least one
endpoint
how close to zero and one the integration reaches
quantile
maps the logit of a cumulative probability to the value of the random variable