feemcorcondia: Core consistency diagnostic for PARAFAC models

In albatross: PARAFAC Analysis of Fluorescence Excitation-Emission Matrices

Core consistency diagnostic for PARAFAC models

Description

Compute the core consistency diagnostic (“CORCONDIA”) by fitting a “Tucker3” core array to the existing PARAFAC loadings.

Usage

  feemcorcondia(
    model, divisor = c("nfac", "core"),
    kind = c('pinv', 'iterative', 'vec'), ...
  )
  ## S3 method for class 'feemcorcondia'
print(x, ...)

Arguments

model	A PARAFAC model returned by feemparafac.
divisor	The divisor used in computation of the CORCONDIA value, see Details.
kind	A string choosing the method used to compute the least squares Tucker3 core. Defaults to “pinv” for PARAFAC models without missing data and “iterative” for models where missing data is present. See Details.
x	An object returned by feemcorcondia.
...	feemcorcondia For kind = 'iterative', forwarded to optim (see Details). Otherwise, not allowed. print.feemcorcondia Ignored.

Details

The “Tucker3” model uses three loading matrices and a small three-way “core array” to describe a larger three-way array:

3{ X_{i,j,k} = \sum_r \sum_s \sum_t A_{i,r} B_{j,s} C_{k,t} G_{r,s,t}

Value

A numeric scalar of class feemcorcondia with the following attributes:

divisor	The divisor argument, expanded to one of the valid options.
core	A three-way array containing the least-squares Tucker core for the given PARAFAC model.

References

\reference

Bro2003-CORCONDIA

See Also

multiway::corcondia

Examples

  data(feems)
  cube <- feemscale(feemscatter(cube, c(20, 14)), na.rm = TRUE)
  # kind = 'vec' is exact but may take a lot of memory
  feemcorcondia(feemparafac(cube, nfac = 3, ctol = 1e-4), kind = 'vec')
  # kind = 'iterative' used by default for models with missing data
  feemcorcondia(feemparafac(cube, nfac = 4, ctol = 1e-4))

albatross documentation built on May 29, 2024, 9:10 a.m.