BridgeChangeRegHybrid: Hybrid Bridge Regression with Change-point

In jongheepark/BridgeChange: Bayesian Change-point models with Bridge prior for analyzing time-series and panel data.

Hybrid Bridge Regression with Change-point

Description

Hybrid Univariate response linear change-point model with Bridge prior.

Usage

BridgeChangeRegHybrid(
  y,
  X,
  n.break = 0,
  scale.data = TRUE,
  intercept = TRUE,
  mcmc = 100,
  burn = 100,
  verbose = 100,
  thin = 1,
  reduce.mcmc = NULL,
  ols.weight = FALSE,
  c0 = 0.1,
  d0 = 0.1,
  nu.shape = 2,
  nu.rate = 2,
  known.alpha = FALSE,
  alpha.start = 1,
  alpha.limit = FALSE,
  alpha.MH = TRUE,
  beta.start = NULL,
  beta.alg = "SVD",
  regime.duration.min = 5,
  waic = FALSE,
  marginal = FALSE
)

Arguments

y	Outcome vector.
X	Design matrix. Columns correspond to variables.
n.break	The number of change-point(s). If n.break = 0, the model corresponds to the usual regression. Default is n.break = 0.
scale.data	If TRUE, y and X are both scaled to have zero mean and unit variance. Default is TRUE. We recommend scale.data = TRUE unless the original data are already scaled.
intercept	If TRUE, estimate intercept by MLE. Default is TRUE. This option does not affect the result when n.break = 0. We recommend intercept = TRUE when the number of break is not zero.
mcmc	The number of iterations for gibbs updates. Default is 100.
burn	The number of burn-in periods for gibbs updates. Default is 100.
verbose	Iterations at which the results are printed on the console. Default is every 100th iteration.
thin	Thinning for gibbs updates. Default is 1 (no thinning).
reduce.mcmc	The number of reduced MCMC iterations for marginal likelihood computations. If reduce.mcmc = NULL, mcmc/thin is used.
ols.weight	If TRUE, OLS estimates are used for adpative lasso weight vector.
c0	Scale parameter for Gamma distribution. Used for the prior of σ^2. Default is 0.1.
d0	Shape parameter for Gamma distribution. Used for the prior of σ^2. Default is 0.1.
nu.shape	Shape parameter for Gamma distribution. Used for the prior for τ. Default is 2.0.
nu.rate	Rate parameter for Gamma distribution. Used for the prior for τ. Default is 2.0.
known.alpha	If TRUE, a user must specify a numeric value [0, 2] in alpha.start. Default is FALSE and therefore α will be estimated.
alpha.start	Starting value for alpha. When known.alpha = TRUE, alpha is fixed to the value of this argument. Default is 1.
alpha.limit	If TRUE, alpha is sampled from (0,1), otherwise alpha is sampled between (0,2). Default is FALSE.
alpha.MH	If TRUE, alpha is updated by the Metropolis–Hastings algorithm. If FALSE the Griddy gibbs sampler is used instead. Default is TRUE.
beta.start	Starting values of beta. If NULL, randomly choose beta.start from OLS or standard normal distribution. Default is NULL.
beta.alg	An algorithm to sample beta. Default is beta.alg = "SVD". Also supported is beta.alg = "BCK" and beta.alg = "CHL".
regime.duration.min	The minimum length of time in each regime. Default is 5. If regime is shorter than this limit, hybrid analysis is skipped.
waic	If TRUE, WAIC is computed after the parameter estimation. Default is FALSE.
marginal	If TRUE, the marginal likelihood is computed based on Chib's method. Default is FALSE.

Value

output

jongheepark/BridgeChange documentation built on Jan. 12, 2023, 4:48 p.m.