betaFeedback

WorksMathematics

orderStatMoments: Exact Means, Variances, and Covariances of Order Statistics in R

Abstract

Calculating the moments, i.e., means, variances, and covariances, of the order statistics of a sample is critical, as they are foundations for order-statistics-based estimation and inference problems. In the early literature, the exact covariance matrices were tabulated only up to the sample size of 50 under normality. Recent software either provides exact calculation only for the means and variances of single order statistics, or obtains the covariances by approximations. We create the R package orderStatMoments to calculate the exact values of these moments for any ranks and any sample size. It covers thirteen built-in distribution families, specified by name and parameters as in R, and allows for continuous distributions defined by the user under relatively weak conditions. It also calculates the moments of any linear combination of order statistics, including sample quantiles, and Lloyd's best linear unbiased estimators with their variances. The work makes two contributions. First, the package provides the exact covariances between order statistics, which no software we found does. Second, we design methods based on the statistical properties of order statistics critical for ultra accurate computing at large sample sizes and various distributions. The accuracy is verified against closed forms at sample sizes up to n=106n = 10^6, against the permutation identities on full matrices, and against 556 cells tabulated in seven publications, where it identified some errors far beyond the stated accuracy. A comparison with five R packages and Python's SciPy shows advantages of this package. Computing is efficient, with a time nearly independent of the sample size.

Paper details

vlnc:W_exsp3gvh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…