Forcasted Panel Data

[crauch]

Hi,
I guess my question is rather simple but I already searched a lot in the Eviews user guide and in this forum but it is quite tricky to enter the right keywords - so, I'm sorry if this has already come up at some point ...

I have very simple panel data for 21 different countries and 2 years (1998 and 2008). For the second year, I cannot use my dependent variable (the reasons are not important for my question), so I use the estimation for 1998 to forecast the values for 2008.

My problem is now that I would like to order my data the following way: first all observations for 1998 and then for 2008 (this is by the way the ordering of the original data in the excel file). Unfortunately, Eviews always orders the data by country making tables and graphs of the forecast output very messy (i.e. it orders in the following way: country1 1998, country1 2008, country2 1998, country2 2008 ... and I would rather prefer: country1 1998, country2 1998, ... country1 2008, country2 2008 ...). So for example, if I open the spread sheet of the forecasted variable, it has this wrong ordering and the 'sort'-button is of no help (it only allows me to start with the last or the first country).

I know, this question is probably very stupid, so I am totally sorry that I may be so blind and cannot see and find the right option to chose.

Thanks a lot four your help!

[EViews Gareth]

Interesting question. I hadn't thought about it before, and I can't think of a way to make it happen either.

Perhaps the best solution is just use the sample to only look at 1998 values or only look at 2008 values.

[EViews Glenn]

For the spreadsheet you can do this by creating a new sorting variable and using that in the display. Something like

Code: Select all

series sortid = @trend*10000+@crossid

The multiplicative factor on trend just has to be large enough so that the trend dominates the cross-section ID (it has to be more than the number of cross-sections). Then in your spreadsheet view, add SORTID to the group and ask to sort by sortid.

That said, we should put in on a list of things to build-in. The interface is a bit tricky, but we'll take a look...

[crauch]

Wow, thanks that was a good way to get around it! The spreadsheet is perfect now, however, is there a possibility to exclude sortid for the graph of the group? Because I think I can show a good relationship between the forecasted values and the actual ones but only if they are sorted the right way and if no sortid variable fills out the whole graph area. I could do this in Excel, too, if this is not possible though ...
Thanks again!

[EViews Glenn]

Interesting. You should note that by default the graph isn't in the sorted form. You can go into the graph and sort using the right mouse button.

But then you are left with the SORTID. Unfortunately, I can't think of a clean way to get rid of it. You could hide it (white on white? delete legend?) and then set the axes to a user scale.