I have a laptop that has 2 cores.
In running my model using Rstan, I realised that some times two chains are delayed.
If I switch to the use of two chains and increase the number of iterations, will the outcome still be reliable?
I have a laptop that has 2 cores.
In running my model using Rstan, I realised that some times two chains are delayed.
If I switch to the use of two chains and increase the number of iterations, will the outcome still be reliable?
Yes, especially if there’s no multi-modality.
We generally recommend running 4 chains. With only two cores you probably also don’t have fast memory, so you might want to test whether running 2 chains in parallel is faster than running them sequentially.
The reason we recommend more chains is to help diagnose problems with multimodality, etc. It’s a kind of robustness check that multiple different starting values of parameters lead to the same result.
Thank you so much Bob_Carpenter.
That response is helpful.
Thank you so much mhollanders, I get it now