Pat: eigenfunctions measured above 0.2 Hz, not much signal below it. Gary agrees.
Vuk: Correction of equation on slide 6.
Daniel: for data points where there is a binary distribution (some retro and some pro), should we try to force the retro solution? This could change the mean and the median.
Pat: not sure, had the same question. Will follow up via email.
Victor/Daniel: if v(f) is a power law, then the coefficients of the biexponential eigenfunctions are independent of frequency.
Pat: velocity at low frequencies is maybe not modelled well with a linear or a power law. Maybe there are better choices?
Victor: usually people don't fit velocity vs frequency with a parametrized curve, instead they use their data (or don't use some frequencies).
Victor/Pat: could try to fix velocities from other estimates, and only estimate other parameters
Vuk: it seems posteriors and priors are similar, not constraining the eigenfunctions more than we already did before the measurements.
Victor: yes, the paper could simply state that the data is consistent with the model.
Vuk: Maybe we can get away with some simpler statistical statement, showing that the model is consistent with the data?
Victor: would be nice, but would not be the selling point of the paper.
Vuk: let's then proceed with what we have. If we come up with a simple statistical test we can add it…