vcov.mhmm: Variance-Covariance Matrix for Coefficients of Covariates of...

In seqHMM: Mixture Hidden Markov Models for Social Sequence Data and Other Multivariate, Multichannel Categorical Time Series

Variance-Covariance Matrix for Coefficients of Covariates of Mixture Hidden Markov Model

Description

Returns the asymptotic covariances matrix of maximum likelihood estimates of the coefficients corresponding to the explanatory variables of the model.

Usage

## S3 method for class 'mhmm'
vcov(object, conditional = TRUE, threads = 1, log_space = FALSE, ...)

Arguments

object	Object of class mhmm.
conditional	If TRUE (default), the standard errors are computed conditional on other model parameters. See details.
threads	Number of threads to use in parallel computing. Default is 1.
log_space	Make computations using log-space instead of scaling for greater numerical stability at cost of decreased computational performance. Default is FALSE.
...	Additional arguments to function jacobian of numDeriv package.

Details

The conditional standard errors are computed using analytical formulas by assuming that the coefficient estimates are not correlated with other model parameter estimates (or that the other parameters are assumed to be fixed). This often underestimates the true standard errors, but is substantially faster approach for preliminary analysis. The non-conditional standard errors are based on the numerical approximation of the full Hessian of the coefficients and the model parameters corresponding to nonzero probabilities. Computing the non-conditional standard errors can be slow for large models as the Jacobian of analytical gradients is computed using finite difference approximation.

Value

Matrix containing the variance-covariance matrix of coefficients.

seqHMM documentation built on July 9, 2023, 6:35 p.m.