* PS9.8, using DATA9-2, for ARCH estimation in Example 9.10 and * Table 9.4 read (data9-2) year r M D genr M1 = M(-1) genr M2 = M(-2) genr D1 = D(-1) genr D2 = D(-2) * suppress first two observations because M2 and D2 are undefined sample 3 36 * estimate model ols r M1 M2 D1 D2 / resid=uhat * generate uhat square and lags genr usq=uhat*uhat genr usq1 = usq(-1) genr usq2 = usq(-2) genr usq3 = usq(-3) * suppress three more observations sample 6 36 * estimate auxiliary regression for ARCH(3) ols usq usq1 usq2 usq3 * compute nrsquared and pvalue gen1 LM1=$n*$r2 distrib LM1 / type=chi df=3 * ARCH(3) is not supported but ARCH(1) term is significant at 9 percent sample 4 36 * try ARCH(1) ols usq usq1 / predict=usqhat gen1 LM2=$n*$r2 * generate predicted variance as observed minus error uhat distrib LM2 / type=chi df=1 * all values of predicted variance are positive print usqhat * compute weights for WLS genr wt=1/usqhat * estimate model by weighted least squares ols r M1 M2 D1 D2 / weight=wt * omit D2 which has an insiginificant coefficient ols r M1 M2 D1 / weight=wt stop