Autoregressive models

Hi, I'm trying to create an auto regressive model, and I want to use AIC to identify the optimal number of lags to include in it. Does anyone know of any code that I can use to do this? I think this requires the econometrics toolbox which I do have. Thanks

Answer 1

[Oleg Komarov]

Example: AR(1)

% Generate series

Series = rand(100,1);

% Set the AR(1) - note the constant variance

Spec = garchset('R',1,'VarianceModel','Constant');

% Estimate

[Coeff,Errors,LLF]  = garchfit(Spec,Series);

% Display

garchdisp(Coeff,Errors)
aicbic(LLF,garchcount(Coeff))

Alternatively:

• Econometrics Toolbox: by James P. LeSage (free): function ar
• MFE_Toolbox (free): function armaxfilter

[Michael]

Hey, thanks for the answer, but would you be able to explain what the lines do? I'm not familiar with the code at all. Thanks!

[Oleg Komarov]

garchset/fit are the functions used to estimate garch/armax models. There's no separate functions for ARMA modelling so you have to set the variance equation in the grach model to constant to estimate ARMAs.

To have an autoregressive of order one - AR(1) - I just specify R (read fucntions doc for more details).

[Michael]

Ok great, thanks, I managed to get sorted!

[Luis Hernandez]

Hi Oleg!! Thank so much for the script!! I've fit an AR(1) model with that script. But, i want to know the meaning of 'K' parameter in coeff set. Is it the value of white noise variance? According to the documentation of garchset the 'K' parameter is just for GARCH models or conditional variance models, why that parameter appears in AR(1) model? Im working with several ARMAX models and need to know the value of that variance. How could find this value? Thanks in advance!!!