In PeppeSaccardi/acmvup: Additive Covariance Modeling via Unconstrained Parametrization

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

acmvup

The goal of acmvup is to provide some useful functions that are implemented following the additive covariance modeling via unconstrained parametrization idea. You will find more detail in my master thesis.

Installation

You can install the released version of acmvup from Github with:

library("devtools")
install_github("PeppeSaccardi/acmvup")

Example

This is a basic example which shows you how to compute the log-likelihood function value, gradient and hessian

library(acmvup)
set.seed(1234)
# Let generate synthetic data, using p = 3
data <- matrix(rnorm(300), ncol=3)

# Then we need the lists containing the covariates for all the
# unconstrained parameters that we want to model
lista_d <- list()
lista_phi <- list()

lista_d[[1]] <- matrix(c(rep(1,100),rnorm(100)), byrow = FALSE, ncol=2)
lista_d[[2]] <- matrix(c(rep(1,100),rnorm(200)), byrow = FALSE, ncol=3)
lista_d[[3]] <- matrix(rep(1,100), byrow = FALSE, ncol=1)

lista_phi[[1]] <- matrix(c(rep(1,100),rnorm(200)),byrow = FALSE, ncol=3)
lista_phi[[2]] <- matrix(rep(1,100),ncol=1)
lista_phi[[3]] <- matrix(c(rep(1,100),rnorm(100)),byrow = FALSE, ncol=2)

Now we are able to use the functions implemented in the package to compute the log-likelihood function value, gradient and hessian, for instance in a suitable random point

par <- rnorm(12)

ll <- optimal_loglik(par,data,lista_phi,lista_d)
gll <- optimal_grad(par,data,lista_phi,lista_d)
hll <- optimal_hessian(par,data,lista_phi,lista_d)

Therefore for our example

print(paste0("The log-likelihood value is ",ll))
print("The gradient vector is ")
gll
print("The hessian matrix is ")
hll

PeppeSaccardi/acmvup documentation built on Dec. 31, 2020, 3:28 p.m.