Benjamin Deonovic

You might want to check why non-positive definite matrices are being formed. Sometimes a matrix is not positive definite in julia due to round off errors. In those cases you...

in the example you have ```julia Gibbs_sigma2 = Sampler([:sigma2], (mu, sigma2, y) -> begin ##sigma2.value is the current value of sigma2 which you are trying to sample a new value...

Your custom sampler doesn't have to be a Gibbs sampler. It can be whatever you want. For MH you would just need to calculate the the acceptance probability for example.

@brian-j-smith in light of transitioning Mamba to Julia 1.0 maybe it would be easier if it excised out the chains logic as well as plotting logic and let development of...

This very vague error occurs when something goes wrong in Mamba. For some reason Julia has stopped propagating the error messages. To get a more accurate assessment of what went...

It looks like for your MWE that you posted you specified lambda as a stochastic array rather than a Stochastic value. ``` julia using Mamba ## Data data = Dict{Symbol,Any}(...

As for the pumps example, what do you mean specifically by remove the alpha and beta parameter? If you completely remove alpha and beta from the model it works just...

It looks like `beta` is not an array but a single value. Perhaps you changed that in the model statement but did not change it in the inits statement. ```...

The MCMC output is just a matrix that you can loop over. Something like this: ``` julia num_iter, num_vars, num_chains = size(sim) Trep = zeros(Float64, num_iter, num_chains) for i in...

After skimming the paper it looks like something reasonable that could be implemented. I won't have time for awhile (December? January?) But feel free to try and take a stab...