Given how consistent the period between the peaks is, can’t you narrow it down to an exact number pretty easily? Even if you fit it the posterior would be almost entirely concentrated around the true number.
I’m trying to start off with simple model from which I add elaborations that increase approximation to real data, which will eventually have more noise, time-varying rates and amplitudes, and a lower-frequency/non-stationary trends on which the pulse is riding.
Oh, and as I said above, the example data I posted is a rather low-noise/low-rate-variability/low-amplitude-variability sample; I have other data that is messier.
Here’s a noisier/less-simple set of real data:
The variability in rate is hard to see, but it’s there and certainly expected from physiology.
For a variable rate, see some discussion and models here http://becs.aalto.fi/~ssarkka/pub/fmri-imm-kfs.pdf
Sorry, no Stan code.
2 Likes