long memory

[Deepika]

Hi
I am looking to estimate long memory and inflation persistence on the lines of Diebold and Rudebusch(1989) "LONG MEMORY AND PERSISTENCE IN AGGREGATE OUTPUT"Journal of Monetary Eronomics 24 (1989) 189-209. can you please help me.

regards
Deepika

[TomDoan]

Isn't that all "off-the-shelf"? Geweke-Porter-Hudak is done using the @GPH procedure. And there's an example in the User's Guide for fitting an ARFIMA model, which includes calculation of the impulse response function. You can either do the two-step estimates that they use or you can estimate the "d" parameter simultaneously with the others.

[Deepika]

Thanks for a prompt reply. But i cannot understand that how the procedure for fitting ARFIMA given in the user guide gives us impulse responses. Kindly help. Thanks in advance.

Regards
Deepika

[TomDoan]

If you don't understand the math (which uses complex and functional analysis), there's not much that I can say. Just trust that it's correct.

[Deepika]

Sorry Sir to put it in a wrong way. what i could not understand is that the example in the user guide uses one AR and one MA lag. Diebold and Rudebusch (1989) have constructed a table for cumulative impulse responses for different combinations of p,q and k-periods where k ranges from 1-400 quarters. i was trying to do that by changing the transfer function. so say for p=2 and q=3, i describe the transfer function as
frml transfer=(3+b*%zlag(t,1))/$
( (2-a*%zlag (t,1) )*%conjg ((1-%z (t,1))^D))
but somehow this is not working. am i right in doing this??
also how will we vary the time period in the program?
Thanks in advance

Regards
Deepika

[TomDoan]

No. The 1 is the 1 as in the lag polynomials (1+bL) and (1-aL). To handle multiple lags, you need to create transfer functions with more than one term with L^j replaced by %zlag(t,j). Before you get too involved in this, did you read their conclusion?