MCMC computation of latent-threshold models.

Estimate a latent-threshold VAR model using a single-move Gibbs sampler as in Nakajima and West (2013)

ltvar(y, p = 2, Intercept = TRUE, nreps = 100, burnin = 10,
  dvb0 = 20, dVb0 = 0.002, dva0 = 2, dVa0 = 0.002, dvh0 = 2,
  dVh0 = 0.002, dm0 = 0, ds0 = 1, da0 = 1, db0 = 1, dg0 = 4,
  dG0 = 0.1, dk0 = 3, nKnots = NULL)

Arguments

y	A TxK matrix with the data
p	number of lags
Intercept	Logical flag whether the model contains an intercept
nreps	number of total mcmc draws
burnin	number of burn-in draws
dvb0, dVb0	prior on volatility of betas
dva0, dVa0	prior on volatiltiy of covariances
dvh0, dVh0	prior on volatility of variances
dm0, ds0	prior on intercept
da0	prior on phi for covariances
db0	prior on phi for betas
dg0, dG0	prior on parameters for stochastic volatility
dk0	latent threshold prior
nKnots	number of blocks in the stochastic volatility sampler

References

Nakajima, J. and M. West (2013) Bayesian Analysis of Latent Threshold Dynamic Models; Journal of Business & Economic Statistics 31 (2), 151-164