[Gretl-users] Time series question (MIDAS multipliers)

Hi all, I have a time-series question. In ordinary ADL models (in LaTeX notation): A(L) y_t = B(L) x_t + u_t the dynamic multipliers are a very well defined concept: they are the coefficients of the polynomial D(L) = B(L)/A(L) (provided of course A(L) is invertible) and it's easy to compute them in hansl via the filter() function or the FFT if you prefer, not to mention the excellent packages "ADMBP" and "lagreg", by Artur and Oleh, respectively. But suppose you have a MIDAS model with an autoregressive part. Something like y_t = \alpha y_{t-1} + \sum_{i=0}^k \beta_i x_{\tau - i} + u_t where $t$ and $\tau$ are the low- and high-frequency time indices, respectively. In this case, you have a model like A(L) y_t + B(L*) x_{\tau} + u_t, where L and L* are two different lag operators, and the relationship between them is far from obvious to me. Does anyone have a reference on how to compute the dynamic multipliers? I have searched a little but found no clues. Can anybody help? ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------