Cross-sectional ranks

[tgoodwin]

I have a pool of time series. I want to create a new set of the series where each observation is the cross-sectional rank (at a point in time) across the pool. How do I do this in e-views? Do I have to use matrix operations? Thanks, Tom

[EViews Gareth]

Put them into a group, convert the group into a matrix (@convert), then use the @colranks function to compute the rankings (probably on the transpose). Something like this:

Code: Select all

matrix data = @convert(mygroup) matrix ranks = @colranks(@transpose(data))

You'll then have to convert the ranks matrix back into a group (mtos)

[tgoodwin]

Thanks Gareth. :)

[tgoodwin]

I spoke too soon. When I use @colranks() I get an Error 500 message. And I can't find this function anywhere in the eviews 7 documentation. Is this a discontinued function?

[EViews Gareth]

make sure your copy of EViews 7 is up to date.

[tgoodwin]

I downloaded an ev7 update and now @colranks works, but there still is nothing in the documentation about it. Is there a @rowranks as well?

[EViews Gareth]

They're sort of hidden functions at the moment (they were added after the release of EViews 7, hence the no-documentation). Rowranks is not yet implemented, which is why you need to use colranks with a transpose. I'll see if I can get rowranks working before the next patch though.