Frequency domain granger causality test

[Gregg orian]

Hi,

Im not sure if this is the right place to post this... but anyway...

I have a time series data set with annual data from year 1996 to 2015. But data from 1997, 1999 and 2001 is missing. I will transpose the time series domain to a frequency domain.

When using time series domain i need the same time interval between each data point. ( example 2002,2003,2004 etc... ) Now the problem is that the interval between differs because of the missing values. ( i have 1996,1998,2000,2002,2003,2004... etc.. ) which, when using time series this creates problem.

My question is, when using frequency domain... is it still a problem that i have different intervals between my data points?

And do anyone have literature on the topic to recommend?

Best regards
Gregg

[Nick Cox]

You don't signal what frequency domain calculations you require but irregular spacing is not an absolute barrier to frequency domain work: there is a mass of literature in astronomy, geophysics, etc. on spectrum calculations for such data. One key word is Lomb-Scargle, fortuitously also least squares (LS) spectrum. I had a crude Stata program somewhere but I never wrote it up.

The bigger deal by far is that series of length 20 strike me as extraordinarily short for such work even without the gaps, but I am happy to label that a personal prejudice not based on much experience.

Oddly as I recall Granger himself worked in the frequency domain early in his career.

[Gregg orian]

Thank you for you answer Nick!

To transpose timeseries domain to Frequency i us fourier analysis tool in excel. I will use your keywords to find some literature on the topic.

I want to do something similar as these guys have done http://www.sciencedirect.com/science...62940816301280

Thanks!