Dan Foreman-Mackey

Excellent. Thanks for this update! This gives me enough to track this down and test it appropriately. I'll let you know if I need more info. Thanks!

Unfortunately this isn't possible with vanilla emcee. You could pin your requirements to `emcee < 3.0`, but there were always issues with the `PTSampler` so I'd probably recommend switching to...

The order is actually correct if you use the (preferred) `get_chain` method instead of the `chain` property. In the back end, the chain is stored as `n_t, n_w, n_d` as...

The only advice I can give in general is what is listed in the message. This isn't a bug, just the reality of MCMC!

What version of emcee are you using? How did you install it? Please make a minimal working example that demonstrates this issue.

Can you please write the minimal working code snippet? It shouldn't depend on your data or your specific model. That way we can nail down where the issue is coming...

Also: you might want to double check the emcee MPI docs: http://emcee.readthedocs.io/en/stable/user/advanced.html#using-mpi-to-distribute-the-computations On Fri, Nov 10, 2017 at 6:31 AM Dan Foreman-Mackey wrote: > Can you please write the minimal...

I'm not too sure what to say here. There isn't any real reason why emcee would cause different behavior with your interpolator - it just directly calls the function you...

There isn't anything fundamentally wrong about this approach (it can actually be useful since you have many independent ensembles in this case). But one thing that you need to be...

The `EnsembleSampler` creates a `RandomState` (that it owns) when it's initialized so the easiest way to have control over the state is to run a `np.random.seed` before initializing the sampler....