Fix layout / plot options of plot(fevd()) function - r

Question

I am currently busy with a Vector Autoregressive Analysis in R, using the vars package. I was wondering if the following things are possible:
1) the previous solution for a better fit on the page worked properly, but now with more variables, we have more plots and the layout is screwed again. I already played with some of the win.graph() parameters, but nothing gives me a proper readable solution.
2) The plot(irf(...) function of the vars package gives a one-graph per page output. I was wondering if this is also possible for the plot(fevd() function by adding some extra (unkown to me) parameters options
3) Also, for readability, I would like to color the graphs, plot(fevd() gives an all-kinds-of-gray output, Is it possible to change those colors?
Thank you in advance
Olivier

Answer 1

You will have to modify the plot function for fevd to do what you want. Here's a modified plot.varfevd function that removes all calls to par(). This allows to use layout properly. The lines that have been removed have been commented out (#). I also removed the parameter that asked for confirmation in "single" plots.
plot.varfevd <-function (x, plot.type = c("multiple", "single"), names = NULL,
main = NULL, col = NULL, ylim = NULL, ylab = NULL, xlab = NULL,
legend = NULL, names.arg = NULL, nc, mar = par("mar"), oma = par("oma"),
addbars = 1, ...)
{
K <- length(x)
ynames <- names(x)
plot.type <- match.arg(plot.type)
if (is.null(names)) {
names <- ynames
}
else {
names <- as.character(names)
if (!(all(names %in% ynames))) {
warning("\nInvalid variable name(s) supplied, using first variable.\n")
names <- ynames[1]
}
}
nv <- length(names)
# op <- par(no.readonly = TRUE)
ifelse(is.null(main), main <- paste("FEVD for", names), main <- rep(main,
nv)[1:nv])
ifelse(is.null(col), col <- gray.colors(K), col <- rep(col,
K)[1:K])
ifelse(is.null(ylab), ylab <- rep("Percentage", nv), ylab <- rep(ylab,
nv)[1:nv])
ifelse(is.null(xlab), xlab <- rep("Horizon", nv), xlab <- rep(xlab,
nv)[1:nv])
ifelse(is.null(ylim), ylim <- c(0, 1), ylim <- ylim)
ifelse(is.null(legend), legend <- ynames, legend <- legend)
if (is.null(names.arg))
names.arg <- c(paste(1:nrow(x[[1]])), rep(NA, addbars))
plotfevd <- function(x, main, col, ylab, xlab, names.arg,
ylim, ...) {
addbars <- as.integer(addbars)
if (addbars > 0) {
hmat <- matrix(0, nrow = K, ncol = addbars)
xvalue <- cbind(t(x), hmat)
barplot(xvalue, main = main, col = col, ylab = ylab,
xlab = xlab, names.arg = names.arg, ylim = ylim,
legend.text = legend, ...)
abline(h = 0)
}
else {
xvalue <- t(x)
barplot(xvalue, main = main, col = col, ylab = ylab,
xlab = xlab, names.arg = names.arg, ylim = ylim,
...)
abline(h = 0)
}
}
if (plot.type == "single") {
# par(mar = mar, oma = oma)
# if (nv > 1)
# par(ask = TRUE)
for (i in 1:nv) {
plotfevd(x = x[[names[i]]], main = main[i], col = col,
ylab = ylab[i], xlab = xlab[i], names.arg = names.arg,
ylim = ylim, ...)
}
}
else if (plot.type == "multiple") {
if (missing(nc)) {
nc <- ifelse(nv > 4, 2, 1)
}
nr <- ceiling(nv/nc)
par(mfcol = c(nr, nc), mar = mar, oma = oma)
for (i in 1:nv) {
plotfevd(x = x[[names[i]]], main = main[i], col = col,
ylab = ylab[i], xlab = xlab[i], names.arg = names.arg,
ylim = ylim, ...)
}
}
# on.exit(par(op))
}
Then, you will need short variable names. Pick acronyms if needed.
library(vars)
data(Canada)
colnames(Canada) <-c("name1","name2","name3","name4")
var <- VAR(Canada , p=4 , type = "both")
Using a wide plot window (using win.graph) and using layout (to get the placement of your eight plots), you can get all charts displayed properly. I also changed the colors of the plots as requested. Finally, we are now using single plots as there are no calls to par() which do not sit well with layout().
win.graph(width=15,height=8)
layout(matrix(1:8,ncol=2))
plot.varfevd(fevd(var, n.ahead = 10 ),plot.type = "single", col=1:4)
plot.varfevd(fevd(var, n.ahead = 10 ),plot.type = "single", col=1:4)