pmoms: The Sample Product Moments: Mean, Standard Deviation, Skew,...

In wasquith/lmomco: L-Moments, Censored L-Moments, Trimmed L-Moments, L-Comoments, and Many Distributions

The Sample Product Moments: Mean, Standard Deviation, Skew, and Excess Kurtosis

Description

Compute the first four sample product moments. Both classical (theoretical and biased) versions and unbiased (nearly) versions are produced. Readers are directed to the References and the source code for implementation details.

Usage

pmoms(x)

Arguments

x	A real value vector.

Value

An R list is returned.

moments	Vector of the product moments: first element is the mean (mean in R), second is standard deviation, and the higher values typically are not used as these are not unbiased moments, but the ratios[3] and ratios[4] are nearly unbiased.
ratios	Vector of the product moment ratios. Second element is the coefficient of variation, ratios[3] is skew, and ratios[4] is kurtosis.
sd	Nearly unbiased standard deviation [well at least unbiased variance (unbiased.sd^2)] computed by R function sd.
umvu.sd	Uniformly-minimum variance unbiased estimator of standard deviation.
skew	Nearly unbiased skew, same as ratios[3].
kurt	Nearly unbiased kurtosis, same as ratios[4].
excesskurt	Excess kurtosis from the Normal distribution: kurt - 3.
classic.sd	Classical (theoretical) definition of standard deviation.
classic.skew	Classical (theoretical) definition of skew.
classic.kurt	Classical (theoretical) definition of kurtosis
classic.excesskurt	Excess classical (theoretical) kurtosis from Normal distribution: classic.kurt - 3.
message	The product moments are confusing in terms of definition because they are not naturally unbiased. This characteristic is different from the L-moments. The author thinks that it is informative to show the biased versions within the “classic” designations. Therefore, this message includes several clarifications of the output.
source	An attribute identifying the computational source (the function name) of the product moments: “pmoms”.

Note

This function is primarily available for gamesmanship with the Pearson Type III distribution as its parameterization in lmomco returns the product moments as the very parameters of that distribution. This of course is like the Normal distribution in which the first two parameters are the first two product moments; the Pearson Type III just adds skew. See the example below. Another reason for having this function in lmomco is that it demonstrates application of unbiased product moments and permits comparisons to the L-moments (see Asquith, 2011; figs. 12.13–12.16).

The umvu.sd is computed by

\hat\sigma' = \frac{\Gamma[(n-1)/2]}{\Gamma(n/2)\sqrt{2}}\sqrt{\sum_{i=1}^{n} (x_i - \hat\mu)^2}\mbox{,}

where \hat\sigma' is the estimate of standard deviation for the sample x of size n, \Gamma(\cdots) is the complete gamma function, and \hat\mu is the arthimetic mean.

Author(s)

W.H. Asquith

References

Asquith, W.H., 2011, Distributional analysis with L-moment statistics using the R environment for statistical computing: Createspace Independent Publishing Platform, ISBN 978–146350841–8.

Hosking, J.R.M., and Wallis, J.R., 1997, Regional frequency analysis—An approach based on L-moments: Cambridge University Press.

Joanes, D.N., Gill, C.A., 1998, Comparing measures of sample skewness and kurtosis: The Statistician, v. 47, no. 1, pp. 183–189.

See Also

lmoms

Examples

# A simple example
PM <- pmoms(rnorm(1000)) # n standard normal values as a fake data set.
cat(c(PM$moments[1],PM$moments[2],PM$ratios[3],PM$ratios[4],"\n"))
# As sample size gets very large the four values returned should be
# 0,1,0,0 by definition of the standard normal distribution.

# A more complex example
para <- vec2par(c(100,500,3),type='pe3') # mean=100, sd=500, skew=3
# The Pearson type III distribution is implemented here such that
# the "parameters" are equal to the mean, standard deviation, and skew.
simDATA <- rlmomco(100,para) # simulate 100 observations
PM <- pmoms(simDATA) # compute the product moments

p.tmp <- c(PM$moments[1],PM$moments[2],PM$ratios[3])
cat(c("Sample P-moments:",p.tmp,"\n"))
# This distribution has considerable variation and large skew. Stability
# of the sample product moments requires LARGE sample sizes (too large
# for a builtin example)

# Continue the example through the L-moments
lmr <- lmoms(simDATA) # compute the L-moments
epara <- parpe3(lmr) # estimate the Pearson III parameters. This is a
# hack to back into comparative estimates of the product moments. This
# can only be done because we know that the parent distribution is a
# Pearson Type III

l.tmp <- c(epara$para[1],epara$para[2],epara$para[3])
cat(c("PearsonIII by L-moments:",l.tmp,"\n"))
# The first values are the means and will be identical and close to 100.
# The second values are the standard deviations and the L-moment to
#   PearsonIII will be closer to 500 than the product moment (this
#   shows the raw power of L-moment based analysis---they work).
# The third values are the skew. Almost certainly the L-moment estimate
#   of skew will be closer to 3 than the product moment.

wasquith/lmomco documentation built on Nov. 13, 2024, 4:53 p.m.