Page 1 of 1
CosmoMC: variable CAMB running time
Posted: February 28 2007
by Paul Hunt
I'm using CosmoMC on a cluster with a modified version of CAMB that runs significantly faster in some regions of parameter space than in others. Will this produce misleading results from CosmoMC? Do I need to ensure that CAMB always takes the same length of time to run whatever the cosmological parameter values?
Thanks in advance,
Paul
CosmoMC: variable CAMB running time
Posted: February 28 2007
by Anze Slosar
Ah, very interesting thought.
It could potentially be a problem if you have one chain exploring "fast" part of parameter space and the other chain exploring "slow" one. However, it shouldn't matter if you ensure that each individual chain is well converged, which it should be, if you require small enough R-1. To be doubly sure, you should make separate plots from individual chains and see that they give compatible results...