Sample numbers plots normalize by ESS

[yannikschaelte]

Add an option to normalize the sample numbers and acceptance rates plot by the ESS.

[Havaldar]

@yannikschaelte I was hoping to tackle this issue and just wanted to make sure no one was already on it. Also would normalizing the sample numbers be equivalent to the relative ESS plot or is it N / ESS? (I am really sorry if this is a silly question)

[yannikschaelte]

Hi @Havaldar , I think no-one is working on this yet -- contributions are very welcome ;) .

For the acceptance rate trajectory, it would be not to plot population_size / n_samples, but ESS / n_samples. This would allow assessing how many simulations were needed per effectively accepted particle, which is something one is often interested in. Similarly, for the sample numbers (stacked bar and time series) plots, it would be not plotting n_samples, but n_samples * population_size / ESS.

However, on second thoughts I think that the latter may be kind of superfluous, as one can see the impact of samples per effective acceptance already in the acceptance rate plot, thus I think only the first one should suffice ...