The VARMAX Procedure

The OUTHT= data set contains predictions of conditional covariance matrices of innovations of the fitted GARCH model that the GARCH statement produces. The following output variables can be created:

The OUTHT= data set contains the values shown in Table 35.10 for a bivariate case.

The OUTHT= data set has the same number of observations as the OUT= data set. Both the OUTHT= and OUT= data sets include any observations at the beginning of the data set that are skipped because of the DIF=, DIFY=, DIFX=, P=, or XLAG= option and include the predicted observations at the end of the data set, which correspond with the LEAD= specification. If you specify an ID statement together with the OUTHT= and OUT= options, then the values of the ID variable in the two data sets correspond with one another.

Consider the following example of the OUTHT= option:

data garch;
   set garch;
   date = intnx( 'month', '01may1972'd, _n_-1 );
   format date yymms.;
run;

proc varmax data=garch;
   id date interval=month;
   model y1 y2 / p=1;
   garch q=1 outht=ht;
   output out=og lead=6;
run;

proc print data=og(obs=8);
   var date y1 for1 std1 lci1 uci1 y2 for2 std2 lci2 uci2;
run;

proc print data=ht(obs=8);
run;

proc print data=og(firstobs=499);
   var date y1 for1 std1 lci1 uci1 y2 for2 std2 lci2 uci2;
run;

proc print data=ht(firstobs=499);
run;

The output in Figure 35.68 and Figure 35.69 shows the first eight observations in the OUT= and OUTHT= data sets, respectively. The first observation is skipped in the GARCH model estimation because of the P=1 option, resulting in the missing values in the first observations in the OUT= and OUTHT= data sets. The output in Figure 35.70 and Figure 35.71 shows the last eight observations in the OUT= and OUTHT= data sets, respectively. In the OUT= data set, the standard deviations of the forecast of dependent variables are time-variant. The last six observations in OUTHT= data set are the multistep forecast of conditional covariance matrices of innovations.