Hi All,

I am new not only on this forum but also on using GAUSS. Thus, I will appreciate any help from anyone.

Right now, I have some questions regarding the output of a GARCH Model that was able to build by looking at an example from GAUSS. However, I have some misunderstanding related to the output.

Below, one can see the output.

Normal Solution AIC -1804.7191 lrs -1818.7191 Coefficients lower cl upper cl beta0[1,1] 0.0551 -0.0692 0.1795 garch[1,1] 1.3969 1.1905 1.6032 garch[2,1] -0.8698 -1.2754 -0.4642 garch[3,1] 0.3965 0.1400 0.6530 arch[1,1] 0.1997 0.0480 0.3515 arch[2,1] -0.1492 -0.3022 0.0038 omega[1,1] 0.0735 0.0035 0.1434 Coefficients lower cl upper cl beta0[1,1] 0.0551053 -0.069243 0.179454 garch[1,1] 1.39688 1.19055 1.60321 garch[2,1] -0.869784 -1.27536 -0.464207 garch[3,1] 0.396494 0.139958 0.653029 arch[1,1] 0.199739 0.047954 0.351524 arch[2,1] -0.149184 -0.302192 0.00382262 omega[1,1] 0.073455 0.0034619 0.143448 Parameters Std. Err T-stat 0.0551053 0.0632884 0.870701 1.39688 0.105014 13.3019 -0.869784 0.206423 -4.2136 0.396494 0.130567 3.03672 0.199739 0.0772526 2.58553 -0.149184 0.0778746 -1.9157 0.073455 0.0356237 2.06197

I was expecting to see the model with the lowest AIC. Please, help me understand what is going here.

Thanks for your understanding

## 3 Answers

0

From your output, it appears that you used the `garchFit`

function from the GAUSS Time Series library to estimate a model with the order of the GARCH parameters set to 3 and the order of the ARCH parameters set to 2.

Your output shows the results of fitting that model. Were you expecting to have all possible combinations of GARCH parameters with an order of 3 or less and ARCH parameters with an order of 2 or less fit and then get the best of those models returned to you?

0

Thanks for your reply.

I would like to fit the data with Garch so that i can choose the one with the smallest AIC among many possible cases

0

You can create two nested loops to iterate over the combinations, like this:

```
max_p = 3;
max_q = 2;
struct garchEstimation out;
// Create a vector to hold all AIC values
aic = zeros(max_p, max_q);
for p(1, max_p, 1);
for q(1, max_q, 1);
// call garchFit here, passing in 'p' and 'q'
// for this iteration, something like:
// out = garchFit(y, X, p, q);
// Collect the AIC values for comparison
aic[p,q] = out.aic;
endfor;
endfor;
```

## Your Answer

## 3 Answers

From your output, it appears that you used the `garchFit`

function from the GAUSS Time Series library to estimate a model with the order of the GARCH parameters set to 3 and the order of the ARCH parameters set to 2.

Your output shows the results of fitting that model. Were you expecting to have all possible combinations of GARCH parameters with an order of 3 or less and ARCH parameters with an order of 2 or less fit and then get the best of those models returned to you?

Thanks for your reply.

I would like to fit the data with Garch so that i can choose the one with the smallest AIC among many possible cases

You can create two nested loops to iterate over the combinations, like this:

```
max_p = 3;
max_q = 2;
struct garchEstimation out;
// Create a vector to hold all AIC values
aic = zeros(max_p, max_q);
for p(1, max_p, 1);
for q(1, max_q, 1);
// call garchFit here, passing in 'p' and 'q'
// for this iteration, something like:
// out = garchFit(y, X, p, q);
// Collect the AIC values for comparison
aic[p,q] = out.aic;
endfor;
endfor;
```