set_lambda: Hyperpriors for Bayesian Models

In bvhar: Bayesian Vector Heterogeneous Autoregressive Modeling

Hyperpriors for Bayesian Models

Description

Set hyperpriors of Bayesian VAR and VHAR models.

Usage

set_lambda(
  mode = 0.2,
  sd = 0.4,
  param = NULL,
  lower = 1e-05,
  upper = 3,
  grid_size = 100L
)

set_psi(shape = 4e-04, scale = 4e-04, lower = 1e-05, upper = 3)

## S3 method for class 'bvharpriorspec'
print(x, digits = max(3L, getOption("digits") - 3L), ...)

is.bvharpriorspec(x)

## S3 method for class 'bvharpriorspec'
knit_print(x, ...)

Arguments

mode	Mode of Gamma distribution. By default, .2.
sd	Standard deviation of Gamma distribution. By default, .4.
param	Shape and rate of Gamma distribution, in the form of c(shape, rate). If specified, ignore mode and sd.
lower	Lower bound for stats::optim(). By default, 1e-5.
upper	Upper bound for stats::optim(). By default, 3.
grid_size	Griddy gibbs grid size for lag scaling
shape	Shape of Inverse Gamma distribution. By default, (.02)^2.
scale	Scale of Inverse Gamma distribution. By default, (.02)^2.
x	Any object
digits	digit option to print
...	not used

Details

In addition to Normal-IW priors set_bvar(), set_bvhar(), and set_weight_bvhar(), these functions give hierarchical structure to the model.

• set_lambda() specifies hyperprior for \lambda (lambda), which is Gamma distribution.

• set_psi() specifies hyperprior for \psi / (\nu_0 - k - 1) = \sigma^2 (sigma), which is Inverse gamma distribution.

The following set of ⁠(mode, sd)⁠ are recommended by Sims and Zha (1998) for set_lambda().

• ⁠(mode = .2, sd = .4)⁠: default

• ⁠(mode = 1, sd = 1)⁠

Giannone et al. (2015) suggested data-based selection for set_psi(). It chooses (0.02)^2 based on its empirical data set.

Value

bvharpriorspec object

References

Giannone, D., Lenza, M., & Primiceri, G. E. (2015). Prior Selection for Vector Autoregressions. Review of Economics and Statistics, 97(2).

Examples

# Hirearchical BVAR specification------------------------
set_bvar(
  sigma = set_psi(shape = 4e-4, scale = 4e-4),
  lambda = set_lambda(mode = .2, sd = .4),
  delta = rep(1, 3),
  eps = 1e-04 # eps = 1e-04
)

bvhar documentation built on April 4, 2025, 5:22 a.m.