icasamp: Sample from Various Source Signal Distributions

In ica: Independent Component Analysis

Sample from Various Source Signal Distributions

Description

Sample observations from the 18 source signal distributions used in Bach and Jordan (2002). Can also return density values and kurtosis for each distribution. Use icaplot to plot distributions.

Usage

icasamp(dname, query = c("rnd","pdf","kur"),
        nsamp = NULL, data = NULL)

Arguments

dname	Distribution name: letter "a" through "r" (see Bach & Jordan, 2002).
query	What to return: query="rnd" for random sample, query="pdf" for density values, and query="kur" for kurtosis.
nsamp	Number of observations to sample. Only used if query="rnd".
data	Data values for density evaluation. Only used if query="pdf".

Details

Inspired by usr_distrib.m from Bach's (2002) kernel-ica MATLAB toolbox.

Value

If query="rnd", returns random sample of size nsamp.

If query="pdf", returns density for input data.

If query="kur", returns kurtosis of distribution.

Author(s)

Nathaniel E. Helwig <[e-mail]>

References

Bach, F.R. (2002). kernel-ica. MATLAB toolbox (ver 1.2) http://www.di.ens.fr/~fbach/kernel-ica/.

Bach, F.R. & Jordan, M.I. (2002). Kernel independent component analysis. Journal of Machine Learning Research, 3, 1-48.

Examples

##########   EXAMPLE   ##########

# sample 1000 observations from distribution "f"
set.seed(123)
mysamp <- icasamp("f","rnd",nsamp=1000)
xr <- range(mysamp)
hist(mysamp,freq=FALSE,ylim=c(0,.8),breaks=sqrt(1000))

# evaluate density of distribution "f"
xseq <- seq(-5,5,length.out=1000)
mypdf <- icasamp("f","pdf",data=xseq)
lines(xseq,mypdf)

# evaluate kurtosis of distribution "f"
icasamp("f","kur")

ica documentation built on July 9, 2022, 1:07 a.m.