我也想知道，特别是有人知道怎样用GMM来估计ARMA-GARCH的参数吗？

用ETS中的自回归语句

PROC IMPORT OUT= ZHP.shuju
            DATAFILE= "C:\Documents and Settings\Administrator\桌面\chuandihanshu.xls"
            DBMS=EXCEL REPLACE;
     RANGE="sheet8$";
     GETNAMES=YES;
     MIXED=NO;
     SCANTEXT=YES;
     USEDATE=YES;
     SCANTIME=YES;
RUN;
ods rtf file="sas235160.rtf";
proc gplot data=zhp.shuju;
symbol1 v=star i=join c=black ;
symbol2 v=none i=r   c=black ;
plot ma5*t=1 ma5*t=2/overlay;
label ma5="上证走势" t="观测值序列号";
run;
ods rtf close;
proc arima  data=zhp.shuju;
identify var=ma5(1) nlag=10;
run;
proc autoreg data=zhp.shuju;
model ma5=t/dw=5 dwprob;
run;
proc autoreg data=zhp.shuju;
model ma5=t/method=ml nlag=5 backstep;
run;
proc autoreg data=zhp.shuju;
model ma5=t/nlag=( 1 2) archtest dwprob;
run;
data zhp.b;
ma5=.;
do t=69 to 73 ;
output;
end;
run;
data zhp.b;
merge zhp.shuju zhp.b;
by t;
run;
proc autoreg data=zhp.b;
model ma5=t/nlag=( 1 2) garch=(q=2,p=2,type=exp);
output out=zhp.p p=yhat pm=ytrend lcl=lcl ucl=ucl;
run;
ods rtf file="f:\sas.rtf";
proc gplot data=zhp.p;
plot ma5*t=1 yhat*t=2 ytrend*t=3
lcl*t=3 ucl*t=3/overlay href=68.5 ;
symbol1 v=star i=join c=black;
symbol2 v=circle i=join c=black;
symbol3 v=none i=join c=black;
label ma5="上证走势" t="观测值序列号";
run;
ods rtf close;