splithalf: Running a Split-Half analysis on a PARAFAC model

In staRdom: PARAFAC Analysis of EEMs from DOM

Running a Split-Half analysis on a PARAFAC model

Description

The samples are split into four subsamples: A,B,C,D. Subsamples are then combined and compared: AB vs. CD, AC vs. BD, AD vs. BC. The results show graphs from the components of each of the 6 models.

Usage

splithalf(
  eem_list,
  comps,
  splits = NA,
  rand = FALSE,
  normalise = TRUE,
  nstart = 20,
  cores = parallel::detectCores(logical = FALSE),
  maxit = 2500,
  ctol = 10^(-7),
  rescale = TRUE,
  strictly_converging = FALSE,
  verbose = FALSE,
  ...
)

Arguments

eem_list	eemlist containing sample data
comps	number of desired components
splits	optional, list of 4 numerical vectors containing the sample numbers for A,B,C and D sample subsets
rand	logical, splits are randomised
normalise	state whether EEM data should be normalised in advance
nstart	number of random starts
cores	number of parallel calculations (e.g. number of physical cores in CPU)
maxit	maximum iterations for PARAFAC algorithm
ctol	Convergence tolerance (R^2 change)
rescale	rescale splithalf models to Fmax, see eempf_rescaleBC
strictly_converging	calculate nstart converging models and take the best. Please see eem_parafac.
verbose	states whether you want additional information during calculation
...	additional parameters that are passed on to parafac

Details

Split data sets can be split suboptimal and cause low TCCs. Therefore, subsamples are recombined in 3 different ways and a TCC close to 1 in only one split combination per component is already a positive result. Check the split sets to check for sample independency.

Value

data frame containing components of the splithalf models

See Also

splithalf_plot, splithalf_tcc

Examples

data(eem_list)

splithalf <- splithalf(eem_list, comps = 6, verbose = TRUE, cores = 2)
splithalf_plot(splithalf)

# Similarity of splits using SSCs
sscs <- splithalf_tcc(splithalf)

staRdom documentation built on July 9, 2023, 5:57 p.m.