R/plot.cvdglars.R

In dglars: Differential Geometric Least Angle Regression

plot.cvdglars <- function(x, ...){
	dev_m <- x$dev_m
	dev_v <- x$dev_v
	nfold <- x$control$nfold
	k <- qt(0.975, nfold - 1)
	dev_up <- dev_m + k * sqrt(dev_v / nfold)
	dev_low <- dev_m - k * sqrt(dev_v / nfold)
	ng <- x$control$ng
	g <- seq(x$g_max, x$g0, length = ng)
	g_hat <- x$g
#	g_hat <- g[which.min(dev_m)]
	df <- sum(abs(x$beta)>0)
	if(!x$family$family %in% c("binomial", "poisson")) df <- df + 1
    plot(g, dev_m, xlab = expression(gamma), ylab = "Deviance",
            ylim = c(min(dev_low), max(dev_up)), pch = 20, type = "n",
            main = "Cross-Validation Deviance", ...)
	segments(x0 = g,y0 = dev_low, y1 = dev_up, col = 8, lty = 2)
	points(g, dev_low, pch = "-")
	points(g, dev_m, pch = 20)
	points(g, dev_up, pch = "-")
	abline(v = g_hat, col = 2, lty = 2, lwd = 2)
	axis(3, at = g_hat, labels = paste("df = ", df, sep = ""), padj = 1)
}