README.md

In EarthSystemDiagnostics/corit: time scale dependent correlation of irregular time series

corit is a package which contains auxiliary functions for estimating time scale dependent correlations of irregularly sampled time series. The functions DirectFiltering, IntegrandInterpolationMethod and InterpolationMethod resample the irregular data to an equidistant spacing before or during filtering the time series. The Pearson correlation of the time series can be estimated using the function CorIrregTimser which allows the filtered time series as an optional output. To test the significance of the estimate, the function CorQuantilesNullHyp provides quantiles of correlations obtained from surrogate records, generated with the function GenNullHypPair as power-law time series with a spectral slope, estimated from the time series by using the function estimateTimserSlopes. The package belongs to the manuscript: Reschke, M., Kunz, T., Laepple, T., 2018. Comparing methods for analysing time scale dependent correlations in irregularly sampled time series data, Computers & Geosciences.

Please contact Maria Reschke ([e-mail]) at the Alfred-Wegener-Institute, Helmholtz Centre for Polar and Marine Research, Germany, for more information.

corit can be installed directly from github:

if (!require("devtools")) {   
  install.packages("devtools")   
}   
devtools::install_github("EarthSystemDiagnostics/corit")

After installation load the package:

library("corit")

Using corit requires the package zoo (available on CRAN) which is automatically installed when installing corit, if it is not yet present. Note: There might be some conflicts when using functions of other packages. In this case, please specific the functions when using.

Code examples to illustrate the application of the R package ‘corit’

#(1) estimating the correlation of a pair of time series using linear interpolation and Gaussian filtering

#Assumes two time series with observations and observation times in a vector used to create a zoo-object.  
library(corit)
time.series1 <- zoo(observations1, order.by = observation.times1)   #create a zoo-object
time.series2 <- zoo(observations2, order.by = observation.times2)
Cor <- CorIrregTimser(
timser1 = time.series1,
timser2 = time.series2,
detr = FALSE,   #remove linear trend time series
method = "InterpolationMethod",
appliedFilter = "gauss",
fc = 1/200, #cut-off frequency
dt = 10,    #time step for the interpolation
int.method = "linear",  #kind of interpolation
filt.output = FALSE)    #return filtered time series 

#(2) applying a significance test for the correlation estimate based on the correlation of independent noise

time.series1 <- zoo(observations1, order.by = observation.times1)
time.series2 <- zoo(observations2, order.by = observation.times2)
slopes <- estimateTimserSlopes( #estimate spectral slopes of the time series
timeseries1 = time.series1,
timeseries2 = time.series2,
int.step = 1)   #time step of the interpolated time series
Quant <- CorQuantilesNullHyp(   #quantiles estimated based on surrogate correlations
timser1 = time.series1,
timser2 = time.series2,
beta.noise1 = slopes$s1,
beta.noise2 = slopes$s2,
detr = FALSE,
rep = 1000, #repetition during Monte Carlo procedure
quant = c(0.05, 0.95),  #quantiles to be estimated
method = "InterpolationMethod",
appliedFilter = "gauss",
fc = 1/200,
dt = 10,
int.method = "linear")

#(3) code of the Monte Carlo procedure for an applied linear interpolation and Gaussian filtering

CorQuantilesNullHyp(timser1, timser2, beta.noise1, beta.noise2, detr, rep, 
quant, method, appliedFilter, fc, dt, int.method) 
{
n <- max(c(max(index(timser1)), max(index(timser2))))
time1 <- index(timser1)
time2 <- index(timser2)
corNullHypPair <- QuantCorPair <- list()
tmpCorNullHypPair <- numeric(length = rep)
for (p in 1:length(fc)) {
for (i in 1:rep) {
#generate surrogate time series and estimate the correlation
tmp <- GenNullHypPair(n, beta.noise1, beta.noise2, time1, time2, detr)
if (method == "InterpolationMethod") {
tmpCorNullHypPair[i] <- cor.test(InterpolationMethod(tmp$y1, 
fc[p], dt, n, int.method, appliedFilter, k), 
InterpolationMethod(tmp$y2, fc[p], dt, n, int.method, 
appliedFilter, k), method = "pearson", alternative = "two.sided", 
na.action = TRUE)$estimate
}
}
corNullHypPair[[p]] <- tmpCorNullHypPair
QuantCorPair[[p]] <- quantile(corNullHypPair[[p]], quant)
}
return(list(corPair = corNullHypPair, Quantile = QuantCorPair))
}

EarthSystemDiagnostics/corit documentation built on May 29, 2019, 1:39 p.m.