Add moments for distributions of order statistics

[sethaxen]

This PR implements moments for OrderStatistic and JointOrderStatistics. These are useful for some statistical tests and for computation of L-moments.

Features

• Exact mean, var, cov, skewness, and kurtosis for order statistics of Uniform and Exponential
• Exact mean and var for order statistics of Logistic
• Exact mean for order statistics of Normal when n<=5
• Numerically integrated marginal moments for all continuous order statistics
• Exact moments for discrete order statistics with finite support and approximate for those with infinite support on the integers.
• Tests

Example

Here's a basic implementation of L-moments:

function lmoment(d::UnivariateDistribution, p::Integer)
    p > 0 || throw(ArgumentError("p must be a positive integer"))
    p == 1 && return mean(d)
    ranks = 1:p
    coefs = @. binomial(p - 1, ranks - 1) * (-1) ^ ranks
    if d isa ContinuousUnivariateDistribution
        means = mean(JointOrderStatistics(d, p, ranks; check_args=false))
    else
        means = mean.(OrderStatistic.(d, p, ranks; check_args=false))
    end
    return (coefs' * means) / p
end

julia> lmoment.(Uniform(0, 10), 1:4) ≈ [5, 10/6, 0, 0]
true

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 292 lines in your changes missing coverage. Please review.

Project coverage is 83.42%. Comparing base (e470421) to head (e8ad400).

Files with missing lines	Patch %	Lines
src/univariate/orderstatistic.jl	0.00%	168 Missing ⚠️
src/multivariate/jointorderstatistics.jl	0.00%	124 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1950      +/-   ##
==========================================
- Coverage   86.19%   83.42%   -2.78%     
==========================================
  Files         146      146              
  Lines        8768     9060     +292     
==========================================
  Hits         7558     7558              
- Misses       1210     1502     +292     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.