gfilinreg: Fiducial sampler for linear regression model

In gfilinreg: Generalized Fiducial Inference for Low-Dimensional Robust Linear Regression

Description

Weighted samples of the fiducial distribution of the parameters of a linear regression model with normal, Student, Cauchy, or logistic error terms.

Usage

gfilinreg(
  formula,
  data = NULL,
  distr = "student",
  df = Inf,
  L = 30L,
  Kmax = 50L,
  nthreads = parallel::detectCores(),
  stopifbig = TRUE
)

Arguments

formula	two-sided formula defining the model
data	dataframe containing the data
distr	the distribution of the error terms, "normal", "student", "cauchy", or "logistic"
df	degrees of freedom of the Student distribution if distr = "student"
L	number of subdivisions of each axis of the hypercube (0,1)^(p+1)
Kmax	maximal number of combinations of indices to use
nthreads	number of threads for parallel computations
stopifbig	logical, whether to stop if the algorithm requires huge matrices

Value

A gfilinreg object, list with the fiducial samples and the weights.

References

Jan Hannig, Randy C.S. Lai, Thomas C.M. Lee. Computational issues of generalized fiducial inference. Computational Statistics and Data Analysis 71 (2014), 849–858. <doi:10.1016/j.csda.2013.03.003>

Examples

set.seed(666L)
x <- c(1, 2, 3, 4)
y <- x + 3 * rcauchy(4L)
gfi <- gfilinreg(y ~ x, distr = "cauchy", L = 30L, nthreads = 2L)
gfiSummary(gfi)

gfilinreg documentation built on March 17, 2021, 1:06 a.m.