MA terms with panel data

[sfawver]

I have 436 regions with 184 observations each I'm trying to forecast. When I include the MA term I get an error saying they're not supported in panel data.

I'm only interested in getting the forecasted values for each each region. Not anything to do with panel data.

Is there a workaround? Like maybe including an id and then adjusting the id by adding observations to it as it loops through the regions?


eviews 7.2

[sfawver]

another file needed for equations to run.

[EViews Gareth]

The only thing you can do is estimate in a non-panel workfile. You could either copy each cross-section to a time-series page one at a time on that page, or you could convert your panel to a pool and estimate each cross-section individually on that pool page.

[sfawver]

is there a way for me to add years to this within the loop?

existing
smpl {%FirstActual} {%LastActual} if newmsa=allmsa(!j)

desired
smpl {%FirstActual}+1yr {%LastActual}+1yr if newmsa=allmsa(!j)

[EViews Gareth]

Use the @dateadd function along with @datestr and @dateval to create a string containing the dates you want, then use those in the smpl command.

[Aijan]

Hi there,

Could you please help me. While estimating a panel model, I came up with an autocorrelation problem. From the ACF and PACF functions I see that the model contains both AR and MA terms. However, I found out that eviews does not allow including MA terms in estimation of panel.

I am using Least Squares, cross-section fixed. Is there a way to take both AR and MA terms into account (maybe through different estimation method or through including some terms that would control for MA).

P.S. I am using EViews 8.
P.S.S. Estimating ARMA model for each cross-section serparately is not a way for me.

Thank you!

[Aijan]

I also found that FGLS can take both heteroscedasticity and autocorrelation problems into account. But found that EViews does not allow estimating FGLS models as well :(