Closed dawe closed 1 year ago
Hi! Thanks for the find! Since num_results
is the number of all samples (burn-in + posterior samples), this is expected because X>Y would mean a negative number of posterior samples. We should catch that before running the sampling process, though.
In our pertpy implementation, we don't have this problem, as it expects the number of posterior samples as an input.
Nice, thanks!
One question, as pertpy
implements CODA and, as far as I can see, the same kind of analysis (and more), is it recommended to switch to it? Are scCODA
and pertpy
being both developed/updated?
As stated in the readme of this repository, pertpy will be the only maintained implementation of the scCODA model. We will not regularly update the scCODA package anymore, and it is recommended to switch to pertpy
The good old RTFM!
Hi all, I've found (by accident) that
model.sample_hmc(num_burnin=X, num_results=Y)
crashes if X > Y: