test.harmonic: Harmonic analysis test for Fast Fourier Transform

In PML: Penalized Multi-Band Learning for Circadian Rhythm Analysis Using Actigraphy

Description

This function conducts harmonic test sequentially based on observations or Fast Fourier Transform (FFT) results.

Usage

test.harmonic(ob, p, fft = FALSE, maxfreq = 10)

Arguments

ob	Either the original observation or FFT results. See parameter fft.
p	The p-value to be considered statistically significant.
fft	If TRUE, ob is FFT results, with the first column frequencies and the second column signals in standardized proportions; if FALSE, ob is a vector of the original observation. The default is FALSE.
maxfreq	To conduct test on at most maxfreq frequencies. The default is 10.

Value

A list of two elements:

sig	The significant frequencies plus the first insignificant frequency.
fft	The FFT results expressed in standardized proportions.

References

Fisher, R. A. (1929). Tests of significance in harmonic analysis. Proceedings of the Royal Society of London. Series A, 125(796), 54-59.

See Also

pharmonic

Examples

data(pa3)

#### test on individuals
ob <- do.call("c",pa3$activity[1:4])
re <- test.harmonic(ob,p=0.05/(length(ob)-1)/2)
re$sig;head(re$fft) ## no harmonic is significant
ob2 <- do.call("c",pa3$activity[11:13])
re2 <- test.harmonic(ob2,p=0.05/(length(ob2)-1)/2)
re2$sig;head(re2$fft) ## 3 significant harmonics

#### test on the population average
re0 <- bandSelect(df=pa3,Nlength=1440*3,Nlambda=100,alpha=1,Ntop=3,
                  cross=FALSE,Ncross=NULL,plot=TRUE)
freq <- data.frame(Frequency=re0$freq,Proportion=colMeans(re0$xprop))
re3 <- test.harmonic(freq,p=0.05/nrow(freq),fft=TRUE)
print(re3$sig,digits=3,row.names=FALSE)

PML documentation built on Feb. 12, 2020, 1:17 a.m.