R/plot.FitAR.R

In FitAR: Subset AR Model Fitting

`plot.FitAR` <-
function(x, terse=TRUE, ...){
res <- resid(x)
n <- length(res)
k <- nrow(x$covHat)
lag.max <- min(n/4,100) 
#Ljung-Box plot and residual acf plot
layout(matrix(c(1,2,1,2),ncol=2))
LBQPlot(res, k=k, lag.max=lag.max)
RacfPlot(x)
if (terse) return(invisible())
#squared residuals
#Ljung-Box plot and residual acf plot
layout(matrix(c(1,2,1,2),ncol=2))
LBQPlot(res, k=k, lag.max=lag.max, SquaredQ=TRUE)
RacfPlot(x, SquaredQ=TRUE)
#normal plot and box plot
layout(matrix(c(1,2,1,2),ncol=2))
ans<-JarqueBeraTest(res)
sti<-paste("Jarque-Bera Test, p-value =",format.pval(ans$pvalue))
#qqnorm(res, ylab="residuals",main=sti)
print(qqmath(~res,ylab="Residual Quantiles",xlab="N(0,1) quantiles",main=sti))
print(bwplot(~res, xlab="residuals"))
#trace of original and bootstrap
layout(matrix(c(1,2,1,2),ncol=2))
y<-x$fits+x$res
TimeSeriesPlot(y,main="Trace of original time series")
zBoot<-Boot(x)
TimeSeriesPlot(zBoot,main="Trace of parametric bootstrap")
#monotonic spread plot
if (length(x$phiHat)>0){
    fits<-as.vector(x$fits)
    fits<-fits-mean(fits)
    plot(fits,sqrt(abs(res)), xlab="fit", ylab="sqrt abs residual")
    title(main="Monotone spread plot")
    lines(lowess(fits,sqrt(abs(res)),f=1),lwd=2,col="blue")
}
#residual/fit plot
print(rfs(x))
#compare theoretical and sample autocorrelations
#layout(matrix(c(1,2,1,2),ncol=2))
layout(matrix(c(1,2,3,1,2,3),ncol=2))
if (length(y)<50)
    LMX<-10
else
    LMX<-40
r<-acf(y, lag.max=LMX, type="correlation",plot=FALSE)$acf
AcfPlot(r[-1], main="Sample ACF")
g<-TacvfAR(x$phiHat, lag.max=LMX)
r<-g[-1]/g[1]
AcfPlot(r, main="Theoretical ACF")
r<-acf(y, lag.max=LMX, type="correlation",plot=FALSE)$acf
AcfPlot(r[-1], main="Sample ACF")
#spectral density plot
PlotARSdf(x$phiHat,InnovationVariance=var(resid(x)),logSdf=TRUE, main="fitted model sdf")
invisible()
}