@erlebach I am confused about the need for .to_event (I’ve described my confusion above as well). I don’t really have much intuition around what .to_event would even mean when applied to draws from a Bernoulli distribution (I can understand better for a normal distribution, because then I think it means that the covariance matrix is something other than the identity matrix). Perhaps it means that underlying algorithms should not rule out some kind of autoregressive dependence? I’ve asked a few related questions in this thread. @fritzo if you have a few minutes, could you chime in?
With respect to the datasets, I just used the CSV links next to the datasets in the book (the links you are referring to) and used Pandas to parse them into a dataframe. I am still working through Chapter 3 at the moment.