rvar: Generate pseudo-random data from a Vector Autoregression
In grayclhn/oosanalysis-R-library: Pseudo out-of-sample forecasting

Description Usage Arguments Details Value Author(s) Examples

This function generates data from a user-specified Vector Autoregression (VAR)

1 2	rvar(nobs, coefficients, intercept, vcv, nburn = 1000, y0 = matrix(0, nlag, neq))

`nobs`	Integer; the number of observations to generate.
`coefficients`	A list of numeric vectors; each vector specifies the coefficients of one of the VAR equations.
`intercept`	A numeric vector containing the intercepts.
`vcv`	A numeric matrix representing the variance-covariance matrix of the innovations.
`nburn`	An integer, the number of draws to discard before generating draws from the VAR.
`y0`	A matrix containing the inital values of the series.

Let β_1 represent the first element of the coefficients list and suppose that there are k equations with l lags. The first equation of the DGP is

y[t,1] = intercept[1] + c(y[t-(1:l),1],…,y[t-(1:l),k]) %*% coefficients[[1]] + epsilon[t,1]

.

A data frame with nobs rows containing the generated series and all of their lags. This is redundant information, but makes it easier to use these results directly in a regression model.

Gray Calhoun gcalhoun@iastate.edu

d <- rvar(10000, list(a = c(0.5, 0, 0.2, 0.1),
                      b = c(0.1, 0.2, 0.5, 0)),
          c(4, 6), diag(2))

lm(a ~ aL1 + aL2 + bL1 + bL2, data = d)
lm(b ~ aL1 + aL2 + bL1 + bL2, data = d)