*
* CUSUM test from p 201
*
open data consump.dat
cal 1950
data(format=prn,org=columns) 1950:1 1993:1
*
rls(csum=cusum,csquared=cusumsq,dfhist=dfs) c / rresids
# constant y
compute rresids(1951:1)= %na
*
* CUSUM test
*
stats(noprint) rresids 1952:1 *
set cusum = cusum/sqrt(%variance)
set upper5 2 44 = .948*sqrt(42)*(1+2.0*(t-2)/42)
set lower5 2 44 = -upper5
graph 3
# cusum
# upper5
# lower5
*
* Post-sample predictive test. The cumulative sum of squares is in the series
* cusumsq. While the numerator could also be written more simply as
* rresids(1993:1)**2, we write it this way as it will generalize easily to more
* than one step.
*
cdf(title="Post-Sample Predictive Test") ftest $
   (cusumsq(1993:1)-cusumsq(1992:1))*(%ndf-1)/cusumsq(1992:1) 1 %ndf-1
*
* Phillips-Harvey Modified Von Neumann ratio
*
diff rresids / drr
compute mvnr=%normsqr(drr)*%ndf/(%normsqr(rresids)*(%ndf-1))
cdf(title="Modified Von Neumann Ratio Test for SC") normal .5*sqrt(%ndf)*(mvnr-2)
*
* Harvey-Collier functional misspecification test. This can be done by just
* computing the sample statistics on the recursive residuals and examining the
* t-stat and significance level provided therein. Note that the result in the
* text is wrong - it was computed by dividing by the square root of T-k, rather
* than multiplying by it.
*
stats rresids 1952:1 1993:1