dbbtest.fit: dbbtest.fit

In vpnsctl/bbreg: Bessel and Beta Regressions via Expectation-Maximization Algorithm for Continuous Bounded Data

Description

Function to run the discrimination test between beta and bessel regressions (DBB).

Usage

dbbtest.fit(
  z,
  x,
  v = NULL,
  link.mean,
  link.precision,
  em_controls = list(maxit = 5000, em_tol = 10^(-5)),
  optim_method = "L-BFGS-B",
  optim_controls = list()
)

Arguments

z	vector of response variables with length n. Each coordinate must belong to the standard unit interval (0,1).
x	matrix of covariates with respect to the mean with dimension (n,nkap).
v	matrix of covariates with respect to the precision parameter. The default is NULL. If not NULL must be of dimension (n,nlam).
link.mean	a string containing the link function for the mean. The possible link functions for the mean are "logit","probit", "cauchit", "cloglog".
link.precision	a string containing the link function the precision parameter. The possible link functions for the precision parameter are "identity", "log", "sqrt".
em_controls	a list containing two elements: maxit that contains the maximum number of iterations of the EM algorithm, the default is set to 5000; em_tol that defines the tolerance value to control the convergence criterion in the EM-algorithm, the default is set to 10^(-5).
optim_method	main optimization algorithm to be used. The available methods are the same as those of optim function. The default is set to "L-BFGS-B".
optim_controls	a list of control arguments to be passed to the optim function in the optimization of the model. For the control options, see the 'Details' in the help of optim for the possible arguments.

Value

Object of class dbbtest, which is a list containing two elements. The 1st one is a table of terms considered in the decision rule of the test; they are sum(z2/n) = sum_i=1^n(z_i^2)/n, sum(quasi_mu) = sum_i=1^n(tildemu_i^2 + tildemu_i(1-tildemu_i)/2) |D_bessel| and |D_beta| as indicated in the main reference. The 2nd term of the list is the name of the selected model (bessel or beta).

See Also

simdata_bes, dbbtest, simdata_bet

Examples

# Illustration using the Weather task data set available in the bbreg package.
n <- 100
x <- cbind(rbinom(n, 1, 0.5), runif(n, -1, 1))
v <- runif(n, -1, 1)
z <- simdata_bes(
  kap = c(1, -1, 0.5), lam = c(0.5, -0.5), x, v,
  repetition = 1, link.mean = "logit", link.precision = "log"
)
z <- unlist(z)
dbbtest.fit(z = z, x = x ,v = v, link.mean = "logit", link.precision = "identity")

vpnsctl/bbreg documentation built on March 14, 2021, 12:11 a.m.