Variance decomposition by frequency

[adam_carrilo]

I am new to DYNARE. I was wondering how to do a variance decomposition by frequency, say 6 to 32 quarters and higher and shorter frequencies. Thanks for the help.

[jpfeifer]

See the manual on conditional_variance_decomposition.

[adam_carrilo]

Thanks Johannes. The manual is the first place I did and there is nothing on variance decomposition by frequency, say business cycles frequency. I know how to do the variance decomposition at various horizons but not by frequency. Any help is appreciated here. Thanks.

[adam_carrilo]

Hi,

I know my question for some people is trivial. I've been looking for an answer for a few hours now and can't find how to do a variance decomposition by frequency. I would really appreciate if someone can help me on this. It would take a few seconds of your time, instead of me looking for hours or days. ANY help much appreciated. Thank you!

[jpfeifer]

What do you mean with "by frequency"? You mean you want e.g. at horizon 8 to decompose the variance in low-frequency and high frequency via a band-pass filter with a particular cutoff? If yes, this is currently not possible. If you specify the hp_filter-option, the conditional_variance_decomposition will take place on hp-filtered data. Other user-defined filters are not supported.

[adam_carrilo]

Johannes,

thank you for your response. By frequency, I meant the variance decomposition for movements between 6 and 32 quarters for example, or in frequency domain from pi/3 to pi/16. I know how to get the variance decomposition at different horizons. Thanks.

[jpfeifer]

Then my comment applies. Only the HP-filter is supported for this. With hp_filter=1600 and conditional_variance_decomposition=[6] will you will at horizon 6 essentially get the variance decomposition of time series passed through a high-pass filter with all frequencies above pi/16 being included in the pass-band (See Baxter/King 1999: MEASURING BUSINESS CYCLES WITH BAND-PASS FILTERS, p. 586-587)

[adam_carrilo]

Thank you Johannes.