problem with estimate ARIMA

Hi !

I try to run such simple code:

Mdl = arima(2,0,2);
EstMdl = estimate(Mdl,train_set,'Display','off');

where "train set" is just vector of some values. For most vectors everything works perfect BUT for some of them I get error

   "The non-seasonal autoregressive polynomial is unstable.'.

Why ? What this error means ? How to fix it ?

Thank you in advance for your help ! :)

Answer 1

[Hang Qian]

Hi Jan,

The error means that there are eigenvalues outside the unit circle. Since an explosive economic time series is unlikely in the real world, the model is not very suitable for the data. You may consider lowering the order of the MA terms, say an ARMA(1,1) might work better.

[Jan Kostrzewa]

Thank you very much :) ARMA(1,1) works better :)

Answer 2

[penny]

hi.then ,how can you predict the later data?