Hodrick-Prescott (HP-filter) as an endogenous entity

[kurtosis]

Hi,
The issue is that i would like to define the HP-filtered time series as an endogenous entity in the model environment.
Say, I have a series y. Then I could calculate y.hpf(lambda=1600) yhp. However, this procedure is only valid outside the model environment.
An alternative way would be to employ matrix algebra, something like yhp=matrixhp*y. But then again, the matrix "matrixhp" is also not defined inside the model environment.
Does somebody have an idea how to solve this?

[atoch]

Bonjour,

I face the same issue. If HP-filter had an implementation as a function it would work.
In the mean time it forces me to have a program to manage it out of the model which is sub-optimal.

Arnaud.

[irma.liem]

And how did you program it then? And how to call it within the model framework?
Thank you for your help.

-irma-

[Karl1]

Is it possible now in the latest version of Eviews to make the hp series a part of the model so that one automatically get the hp-series also for the forecasts and not only the historical data?

[EViews Gareth]

No.

[xprimexinverse]

Not sure I fully understand what the problem is here, but one idea would be to consider the fact that a HP filter can be cast into state-space form. Treating the HP filter as a special case state-space model (with lambda entering via restrictions on the variance terms) you could then merge a state-space object into a model object thereby endogenizing the HP filter. Does that sound like a solution to what you're trying to do? If so, I can attach a state-space variation of the HP filter. As I said, not sure exactly what you're trying to do, so let me know.

Cheers,
Graeme

[trubador]

Casting state space form of HP filter will not work either. As far as know, model object cannot endogenize latent variables (i.e. states). However, building the whole model within the state space framework might work.