favar: Factor-Augmented Vector Autoregression
In joergrieger/bvar: Estimation and forecasting of bayesian VAR models

Description Usage Arguments Details Value References

Function to estimate factor-augmented vector autoregressions using a 2-step procedure.

1 2	favar(data, priorObj, factordata, nreps, burnin, alpha, beta, tau2, c2, li_prvar, priorm, stabletest = TRUE, nthin = 1)

`data`	data that is not going to be reduced to factors
`priorObj`	An S3 object containing information about the prior.
`factordata`	data that is going to be reduced to its factors
`nreps`	total number of draws
`burnin`	number of burn-in draws.
`alpha, beta`	prior on the variance of the measurement equation
`tau2`	variance of the coefficients in the measurement equation (only used if priorm=2)
`c2`	factor for the variance of the coefficients (only used if priorm=2)
`li_prvar`	prior on variance of coefficients (only used if priorm = 1)
`priorm`	Selects the prior on the measurement equation, 1=Normal-Gamma Prior and 2=SSVS prior.
`stabletest`	boolean, check if a draw is stationary or not
`nthin`	thinning parameter

Estimates a favar-model using a 2-step procedure. In the first step the factors are extracted from the series using principal components. In the second step, a VAR-model of order p of both the factor series and other variables is estimated. To uniquely identify the favar all series are normalized with mean 0 and standard deviation of 1. Furthermore, the variance-covariance matrix is assumed to be diagonal. The VAR-model is of the form

≤ft[\begin{array}{c}Y_t\\ F_t\end{array}\right]=Φ(L)≤ft[\begin{array}{c}Y_{t-1}\\ F_{t-1}\end{array}\right]+w_t

and the observation equation takes the form

X_t=Λ^fF_t+Λ^yY_t+e_t

Since a model with \tilde{Λ}^f=Λ^fH and \tilde{F}_t=H^{-1}F_t are observationally equivalent to eqn\Lambda,F we impose the standard normalization restriction implicit in the principal components. One interpretation of the factors is that they are a diffusion index as in Stock and Watson (1998)

A S3-object of the class favar.

Bernanke, Ben S., Jean Boivin and Piotr Eliasz, Measuring the effects of monetary policy: a factor-augmented vector autoregressive (favar) approach

Stock, James and Mark Watson, Diffusion Indexes, NBER Working Paper No. 6702, 1998

joergrieger/bvar documentation built on July 3, 2020, 5:34 p.m.