nulleigs: Find the parameters that contribute to a singular covariance...

In jkbest2/spatq: Use spatially varying catchability to standardize fishery-dependent data

Find the parameters that contribute to a singular covariance matrix

Description

Take the eigendecomposition of the fixed-effect covariance matrix, filter out the smallest eigenvalues, and then order the parameter names by the loadings in the corresponding eigenvectors.

Usage

nulleigs(x, reltol = 1e-04, abstol = NULL, vectrans = abs)

fixpar_nulleigs(sdr, reltol = 1e-04, abstol = NULL)

jointprec_nulleigs(sdr, reltol = 1e-04, abstol = NULL)

Arguments

x	A matrix
reltol	Relative tolerance for small eigenvalues
abstol	Absolute tolerance for small eigenvalues; takes precedence
vectrans	Transformation function applied before sorting eigenvector elements, usually either abs or identity
sdr	An sdreport object from a TMB model

Details

The reltol argument sets a minimum for the ratio of each eigenvalue to the largest eigenvalue. The abstol argument sets an absolute minimum value. The latter takes precedence if it is passed.

Value

A list with values (eigenvalues), vectors (eigenvectors), and sortedpars, which is a matrix of parameter names with each column ordered by the corresponding loadings of the eigenvectors (largest loadings at the top)

Functions

• fixpar_nulleigs: Find (nearly) null eigenvalues in the fixed parameter covariance matrix

• jointprec_nulleigs: Find (nearly) null eigenvalues in the joint precision matrix

Author(s)

John K Best

jkbest2/spatq documentation built on Sept. 22, 2022, 3:22 a.m.