LOG(GARCH) = C(4) + C(5)*ABS(RESID(-1)/@SQRT(GARCH(-1))) +
C(6)*RESID(-1)/@SQRT(GARCH(-1)) + C(7)*LOG(GARCH(-1))
Variable Coefficient Std. Error z-Statistic Prob.
C 0.000633 0.000116 5.474713 0.0000
AR(1) 0.992731 0.003498 283.7720 0.0000
MA(1) -0.996871 0.001891 -527.2477 0.0000
Variance Equation
C(4) -0.643502 0.062172 -10.35042 0.0000
C(5) 0.288703 0.020650 13.98095 0.0000
C(6) -0.139497 0.011704 -11.91867 0.0000
C(7) 0.954554 0.005708 167.2299 0.0000
R Output
Robust Standard Errors:
Estimate Std. Error t value Pr(>|t|)
mu 0.000721 0.000079 9.1181 0.000000
ar1 0.535674 0.010577 50.6463 0.000000
ma1 -0.588905 0.014082 -41.8185 0.000000
omega -0.415480 0.080841 -5.1395 0.000000
alpha1 -0.131048 0.038709 -3.3854 0.000711
beta1 0.954220 0.008874 107.5255 0.000000
gamma1 0.318389 0.054693 5.8214 0.000000
An EGARCH model's gamma coefficient is positive in R but negative in EViews, suggests a difference in the package’s estimation methods or assumptions. Technically, the differences may result from software parameterization, assumptions, optimisation techniques, etc.
In real sense, without data and attribute specifics, choosing between the two options is difficult. You can still use AIC, BIC, and LR-tests in both software packages to compare estimates by analysing the importance of the estimated coefficients and evaluating model fit statistics.
Generally, EViews can also offer EGARCH model estimate methods with different parameterizations and assumptions than R. Thus, EViews can estimate better by considering data characteristics.
In that angle, choosing an estimate in your regard requires understanding the packages methods, assumptions, fit statistics, and coefficient significance.
Dickson Utonga Thanks for the answer I was also presuming the same logic, however was unable to put same parameters in both to check
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India