I have done a time-domain decomposition of a generalized forecast error variance from a VAR model of exchange rates and inflation rates. The data are monthly. I am not very adept at doing the frequency domain counterpart of this decomposition but want to do it in order to obtain different frequency-band responses to variable shocks hitting the system and comment on causality in the various frequency-bands. I think I got completely bogged down (not only) in the Nyquist problem (i.e. highest detectable frequency). Specifically, suppose you look at the high-frequency band from $\pi/2$ to $\pi$. What response times are you actually looking at? I think it is roughly two to eight months. Am I right?
2
-
$\begingroup$ Hi: Since you haven't received any responses, it might be better to ask this question dsp.stackexchange or math.stackexchange. Or, if you want to learn dsp in general, go over to dsp.stackexchange and lurk over there. I did that for a while and it helped me to become a beginner. Also, there are many good intro dsp books. I wouldn't want to start saying which ones because everyone is different. They have recommendations over there also. $\endgroup$– mark leedsCommented Apr 3 at 21:02
-
$\begingroup$ @markleeds Thank you. I might do that. $\endgroup$– Pavel FilipCommented Apr 4 at 7:19
Add a comment
|