Ensemble correlation

Source: R/correlation.regression.functions.R

Primary function for calculating correlation ensembles

corEns(
  time.1,
  values.1,
  time.2,
  values.2,
  bin.vec = NA,
  bin.step = NA,
  bin.fun = mean,
  max.ens = NA,
  percentiles = c(0.025, 0.25, 0.5, 0.75, 0.975),
  min.obs = 10,
  fdr.qlevel = 0.05,
  gaussianize = TRUE,
  ...
)

Arguments

time.1

matrix of age/time ensembles, or single column

values.1

matrix of values ensembles, or single column

time.2

matrix of age/time ensembles, or single column

values.2

matrix of values ensembles, or single column

bin.vec

vector of bin edges for binning step

bin.step

spacing of bins, used to build bin step

bin.fun

function to use during binning (mean, sd, and sum all work)

max.ens

Maximum number of ensembles to use

percentiles

quantiles to calculate for regression parameters

min.obs

minimum number of points required to calculate regression

fdr.qlevel

target false discovery rate (most users won't want to change this)

gaussianize

Convert data to Gaussian distribution before correlating?

...

Arguments passed on to corMatrix

ens.1

matrix of age-uncertain columns to correlate and calculate p-values

ens.2

matrix of age-uncertain columns to correlate and calculate p-values

isospectral

estimate significance using the Ebisuzaki method (default = TRUE)

isopersistent

estimate significance using the isopersistence method (default = FALSE)

p.ens

number of ensemble members to use for isospectral and/or isopersistent methods (default = 100)

cor.method

correlation method to pass to cor() "pearson" (default), "kendall", or "spearman". Note that because the standard Student's T-test for significance is inappropriate for Kendall's Tau correlations, the raw and effective-N significance estimates will be NA when using "kendall"

Value

list of ensemble output and percentile information

Author

Nick McKay