-3
$\begingroup$

what is the most appropriate software for building an ARMA forecasting model? EViews, Minitab,...? Best, Milos

$\endgroup$
2
  • 2
    $\begingroup$ Define criteria for what is appropriate. In its current form the question is not answerable, since it does not have a definite answer. See the FAQ. $\endgroup$
    – mpiktas
    Sep 16, 2011 at 11:25
  • $\begingroup$ Do you know any statistical software already? Do you have any experience with ARIMA and variants? Do you have a budget to buy something, or would you be willing to use open source? Do you want a GUI, or are you comfortable typing commands? How large is your data source, and in what format? What kind of data is this? How are you preprocessing/checking the data (i.e. is the software ONLY to do ARMA, or does it need to handle the entire process)? $\endgroup$
    – Wayne
    Sep 16, 2011 at 15:17

1 Answer 1

2
$\begingroup$

As mpiktas points out one has to specify the criteria. If one is concerned about adhering to fundamental mathematical/statistical principles you might look at the book ( page 671 and on) Principles of Forecasting. If you want to get a recent evaluation , incomplete in my opinion, you might see this software survey. Sometimes ease-of-use is important , sometimes detecting and incorporating Gaussian violations and augmenting the simple ARMA model is important. Sometimes the degree of "automatic model selection" is important and how they limit the "sample space" from wich to select the "best model". There have been a lot of developments in ARMA modelling and most pieces of software seriously lag academic developments. If I were you I would take one time series that you were interested in and evaluate the software vendors on that one series, not necessarily for how well their automatic procedure worked but HOW IT WORKED. In other words understand their model selection/validation procedure.

$\endgroup$
0

Not the answer you're looking for? Browse other questions tagged or ask your own question.