I was wondering if someone could tell me or point to resources which discuss what kind of of windowing corrections factors one should use when using a frequency domain Whittle likelihood for spectral estimation (or frequency domain parameter estimation in general).

I know that when using FFTs to estimate PSD, the square of the absolute value of FFTs is scaled by $\frac{2}{w^2}\frac{1}{f_s T}$, where w is the window function. Should the same scaling factor be used in the Whittle likelihood too? I tried this scaling with simple spectral models with a single parameter and even though the spectra match what I expect, the likelihood estimates for the spectral parameter do not match what I expect.

enter image description here

enter image description here

The "data" is colored gaussian data simulated using the correct spectral values. I have tried a handful of simple models, and I never had good consistent recoveries. I have been trying to figure this out for a few weeks with little progress and lot of frustration. Any help will be greatly appreciated. Thank you!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.