fdnonpar-package: The fdnonpar package

In fdnonpar: Methods for Multivariate or Functional Data

Description

The package fdnonpar is a collection of tools for visualisation and nonparametric analysis of grouped data that consist of an independent (x-) variable and a dependent (y-) variable. Such data could be time series or values of a function, or generally any multivariate data. In the latter case, x is just an index variable.

The proposed tests do not take the x-variable into account. Since the x-values are usually equidistant, this does not really matter. In the current version (0.3), it is assumed that the data samples to be compared share the same (equidistant) x-variables.

Data classes, visualisation of data

fdsample	basic data type, a sample of curves
fdenvelope	two boundary curves of an envelope, a specialized fdsample object
print.fdsample	print brief details of an fdsample
plot.fdsample	plot the individual members of an fdsample
plot.fdenvelope	plot an envelope object
summaryplot	plot individual curves, envelopes or summary functions
	such as the mean of an fdsample objectr

Manipulation and summary functions

[.fdsample	extract or replace curves form a sample
apply.fdsample	apply a summary function to the function values
mean.fdsample	mean of the function values
median.fdsample	median of the function values
quantile.fdsample	quantiles of the function values
pwEnvelope	pointwise envelope

Tests

tL2.permtest	Comparison of two groups,
	using square integrated Welch-t-statistic
rankEnv.test	Rank envelope test: compare a single observation to a group (e.g. simulated data)
rankCount.test	Rank count test: refined p-value for rank envelope test

Author(s)

Ute Hahn, [e-mail]

References

Hahn, U. (2012) A studentized permutation test for the comparison of spatial point patterns. Journal of the American Statistical Association 107 (498), 754–764.

M. Myllymaki, T. Mrkvicka, P. Grabarnik, H. Seijo and Ute Hahn (2015) Global envelope tests for spatial processes, http://arxiv.org/abs/1307.0239v3.

Examples

# simulated data sets, consisting of 8 and 9
x <- seq(0, 1, .1)
y1 <- replicate(8, rnorm(length(x), mean = x, sd = .2))
y2 <- replicate(7, rnorm(length(x), mean = x*1.3, sd = .2))
y1b <- replicate(7, rnorm(length(x), mean = x, sd = .2))

xy1 <- fdsample(x, y1)
xy2 <- fdsample(x, y2)
xy1b <- fdsample(x, y1b)

# visualize the data sets
# require plutils which contains the generic to "summaryplot"
require(plutils)

summaryplot(xy1, envprob=1)
summaryplot(xy1b, add = TRUE, col = "blue")
summaryplot(xy2, add = TRUE, col = "red")

# there should be significant difference between xyl1
# and xyl2, but not between xyl1 and xyl1b.
# However, with simulated data, everything is possible...

tL2.permtest(xy1, xy2)
tL2.permtest(xy1, xy1b)

fdnonpar documentation built on May 2, 2019, 5:54 p.m.