How can I apply the AIC criterion to time series data using MATLAB?

Question

As part of the Econometrics Toolbox, how can i apply the AIC criterion to time series data using MATLAB.

Is there any way without using GARCH functions such as garchfit. If the only way to AIC is to apply GARCH functions, what is meant by number of parameters??

Answer 1

Eventhough I am not familiar with GARCH models, here's an example of how to use AIC/BIC for model selection (based on examples in the documentation):

load Data_MarkPound
dem2gbp = price2ret(Data);

%# fit model with specification parameters spec1
spec1 = garchset('P',1, 'Q',1, 'Display','off');
[coeff1,errors1,LLF1] = garchfit(spec1, dem2gbp);
numParams1 = garchcount(coeff1);
%#garchdisp(coeff1,errors1)

%# fit model with specification parameters spec2
spec2 = garchset('P',2, 'Q',1, 'Display','off');
[coeff2,errors2,LLF2] = garchfit(spec2, dem2gbp);
numParams2 = garchcount(coeff2);
%#garchdisp(coeff2,errors2)

%# find the best model with the smallest AIC/BIC
numObsv = length(dem2gbp);
[AIC1,BIC1] = aicbic(LLF1, numParams1, numObsv)
[AIC2,BIC2] = aicbic(LLF2, numParams2, numObsv)