CosmoCoffee

[charles jose]

Hi
I am a new user of COSMOMC.

In COSMOMC 'params.ini' file we need to specify "MPI_Converge_Stop" to decide when to stop the chain. If I understood correctly, 'COSMOMC readme' says this number needs to be less than 0.2. The default value of this number in 'params.ini' file is 0.03. I used this number and produced chains. But when I analyzed the chains using 'getdist' I got the following message
"RL: Not enough samples to estimate convergence stats"

This problem has been solved after I put "MPI_Converge_Stop = 0.01".

I would like read about this convergence criteria. The references given in "COSMOMC readme" are invalid. So can somebody help me with some documentation on the statistics (R statistic) that determines the convergence of the chains.

Thanks in advance

Charles

[charles jose]

The reason I am asking this question is the following.

I myself have added a new module to COSMOMC which returns a new likelihood (like HST or BAO) for some new data. This new likelihood will be added up to likelihood from WMAP data alone. This is for constraining cosmological parameters using WMAP and this new data. The program is running fine and the new likelihood is produced properly. But the COSMOMC code is not stopping and seems like running for ever. When I analyze using "getdist" I am getting the message.
"RL: Not enough samples to estimate convergence stats"

I also found that the likelihood function is rising sharply from a global minimum value with variation of cosmological parameters. So I guess that there are not many samples to determine the statistics.

Some documentation on this R−1 statistics will be really helpful.

Thanks again

Charles

[Alan Zablocki]

I too have found that the links did not work.

[Antony Lewis]

Thanks, I fixed the links in the readme.