Statistics and Algorithms / GARCH Models / GARCH Models (Univariate) /

UV GARCH Output

The output from a GARCH instruction is similar to that from a LINREG, but with many fewer summary statistics; for instance, no \(R^2\) or anything based upon sums of squared residuals, since it isn't minimizing that.

GARCH Model - Estimation by BFGS

Convergence in 18 Iterations. Final criterion was 0.0000093 <= 0.0000100

Dependent Variable DLOGDM

Usable Observations 1866

Log Likelihood -2068.1265

Variable Coeff Std Error T-Stat Signif

********************************************************************************

1. Mean(DLOGDM) -0.020635802 0.015767968 -1.30872 0.19063035

2. C 0.016180175 0.005181947 3.12241 0.00179375

3. A 0.110125419 0.015859223 6.94394 0.00000000

4. B 0.868369919 0.018692362 46.45587 0.00000000

The coefficients are listed in the order:

1.Mean model parameters in the standard order for regressions. If you use the default mean model, the coefficient will be labeled as Mean(series). If you used the REGRESSORS option, these will be labeled as they would for any other regression.

2.Constant in the variance equation (labeled as C).

3.“ARCH” (lagged squared residuals) parameters, in increasing order of lag (labeled as A, or A(lag) if you used more than one lag)

4.“GARCH” (lagged variance) parameters, if any, in increasing order of lag (labeled as B or B(lag))

5.Asymmetry coefficients if any (labeled as D or D(lag))

6.XREGRESSORS variables, labeled as they would in any other regression.

This is also the order in which the coefficients appear in the %BETA vector defined by GARCH, and is the order for the INITIAL vector, if you want to use that.