read_pars {R2admb} | R Documentation |
Read in parameters from an AD Model Builder run
Description
Reads coefficients, standard errors, log-likelihoods, maximum gradients, correlation and variance-covariance matrices from AD Model Builder output files
Usage
read_pars(fn, drop_phase = TRUE, covfn = "admodel.cov",
warn_nonstd_rep = TRUE)
read_psv(fn, names = NULL)
read_rep(fn, names = NULL, warn_nonstd_rep = TRUE)
Arguments
fn |
(character) Base name of AD Model Builder |
drop_phase |
(logical) drop negative-phase (fixed) parameters from results? |
covfn |
(character) file name for covariance matrix information |
warn_nonstd_rep |
warn if report file is in nonstandard format? |
names |
(character) Names of variables |
Details
Given the output from an ADMB run on FOO.tpl, read_pars
reads the
files FOO.par (parameters, log-likelihood, max gradient); FOO.std (standard
deviations); FOO.cor (correlations); FOO.rep (report variables);
admodel.hes for hessian; and admodel.cov for
covariance matrix. read_psv
reads the output of MCMC runs.
read_rep
(called by read_admb
to read the .rep
file) first
checks if the report file is in a standard format: first line starts with a
comment character (#); thereafter, each block starts with a single commented
line containing the name of the parameter (possibly ending with a colon),
followed by a block of all-numeric lines, which are read as a single vector.
If the report file is in a standard format, the values are added to the end
of the coefficients list. Otherwise, the numeric values from the report file
are included in the results as a single, concantenated numeric vector.
Value
List containing the following elements
coefficientsparameter estimates
coeflistparameter estimates in list format, with proper shape (vectors, matrices, etc.)
seestimated standard errors of coefficients
logliklog-likelihood
maxgradmaximum gradient of log-likelihood surface
corcorrelation matrix
vcovvariance-covariance matrix
nparnumber of parameters
heshessian matrix (only if no vcov matrix)
reportvalues from report file (if non-standard report file)
Warnings
The
coeflist
component is untested for data structures more complicated than scalars, vectors or matrices (i.e. higher-dimensional or ragged arrays)Because ADMB hard-codes the file name for covariance matrix information (
admodel.cov
), care is necessary when running different models in the same directory; users may want to rename this file by hand and use thecovfn
argument