Library: | VaR |
See also: | VaRest VaRpred VaRdiagplot VaRdiagtable VaRtimeplot |
Quantlet: | VaRgrdiag | |
Description: | produces calibration and discrimination plots which verify validity of probability forecasts. |
Usage: | {freq,calibr,discr}=VaRgrdiag(int,sig,real,g) | |
Input: | ||
int | mx2 matrix containing intervals in the rows. | |
sig | nx1 vector of predicted standard deviations (assume normality) | |
real | nx1 vector of realizations. | |
g | scalar, number of grid points. | |
Output: | ||
freq | graphical object containing rel. frequencies of the probability forecasts calculated on the given number of grid points. | |
calibr | graphical object containing the calibration plot. | |
discr | graphical object containing the discrimination plot. |
library("VaR") x=read("kupfer") ; time series x=x[1:1001] y=diff(log(x)) ; returns sig=VaRest(y)[,2]/qfn(0.99) sig2=VaRest(y,"EMA")[,2]/qfn(0.99) y=y[251:1000] intervals=(-0.01~0.01)|(-0.017~0.017)|(-0.005~0.005)|(-0.002~0.002) {fr1,ca1,di1}=VaRgrdiag(intervals,sig,y,5) {fr2,ca2,di2}=VaRgrdiag(intervals,sig2,y,5) disp=createdisplay(2,3) show(disp,1,1,fr1) show(disp,1,2,ca1) show(disp,1,3,di1) show(disp,2,1,fr2) show(disp,2,2,ca2) show(disp,2,3,di2)
Graphics comparing two methods of VaR prediction from the "probability forecasts" point of view.