| 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.