pop_pred_samp: generate population prediction sample from parameters

Description

This [EXPERIMENTAL] function combines several sampling tricks to compute a version of an importance sample (based on flat priors) for the parameters.

Usage

pop_pred_samp(
  object,
  n = 1000,
  n_imp = n * 10,
  return_wts = FALSE,
  impsamp = FALSE,
  PDify = FALSE,
  PDmethod = NULL,
  Sigma = vcov(object),
  tol = 1e-06,
  return_all = FALSE,
  rmvnorm_method = c("mvtnorm", "MASS"),
  fix_params = NULL,
  ...
)

Arguments

object: a fitted mle2 object
n: number of samples to return
n_imp: number of total samples from which to draw, if doing importance sampling
return_wts: return a column giving the weights of the samples, for use in weighted summaries?
impsamp: subsample values (with replacement) based on their weights?
PDify: use Gill and King generalized-inverse procedure to correct non-positive-definite variance-covariance matrix if necessary?
PDmethod: method for fixing non-positive-definite covariance matrices
tol: tolerance for detecting small eigenvalues
return_all: return a matrix including all values, and weights (rather than taking a sample)
rmvnorm_method: package to use for generating MVN samples
fix_params: parameters to fix (in addition to parameters that were fixed during estimation)
Sigma: covariance matrix for sampling
...: additional parameters to pass to the negative log-likelihood function

References

Gill, Jeff, and Gary King. "What to Do When Your Hessian Is Not Invertible: Alternatives to Model Respecification in Nonlinear Estimation." Sociological Methods & Research 33, no. 1 (2004): 54-87. Lande, Russ and Steinar Engen and Bernt-Erik Saether, Stochastic Population Dynamics in Ecology and Conservation. Oxford University Press, 2003.