A question regarding the fitting range.

[zamanzadeh98]

Hey,
Thank you for making this package and contributing to open source in neuroscience.
I have a question regarding the fitting range in PYRASA and would appreciate your assistance.

I aim to analyse the 2-40 Hz range in power spectra and calculate theta, alpha, beta, and low gamma frequency bands. However, given my understanding of this paper (Gerster et al. 2022), the evaluated range is larger than the fitting range: f_eval_max = f_fit_max × h_max (in my case, 40 × 1.9 ≈ 76 Hz).
According to the paper's recommendations, I should set the evaluated range (not the fitting range) to my frequency range of interest. This would mean setting my fitting range to approximately f_fit_max = 40 / 1.9 ≈ 21 Hz, so that f_eval_max = 40 Hz. However, doing so would result in power spectra that do not cover the full 2-40 Hz frequency range I need for calculating low gamma or beta bands.
Does PYRASA handle this automatically, or should I manually adjust my fitting range? If I need to adjust it, how can I still extract oscillatory power across the full 2-40 Hz range?
I would appreciate your assistance in choosing the right fitting range for my analysis.

Best wishes,

[schmidtfa]

Hi,
Thanks for using pyrasa :)
Your understanding is correct IRASA is evaluating a larger frequency range and there are some pitfalls associated with it that are nicely described in the Gerster et al. 2022 paper. In pyrasa I therefore introduced some safeguards to avoid evaluating frequency ranges that are too high or too low e.g. Nyquist or high/lowpass filters, but to take advantage of all of them you need to pass some information about your filter settings to the model (see here).
When it comes to choosing the right fitting range I would generally recommend to evaluate a larger frequency range using the irasa model. Afterwards you can use the .get_peaks() method to extract peaks in the theta, alpha, beta, and low gamma frequency bands (see here for the parameters you can set) or the .fit_aperiodic_model() method (see here for the parameters you can set). Hope that helps :)
Let me know if you got further questions