Cosmomc with mock data

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
joby pk
Posts: 12
Joined: February 26 2016
Affiliation: Asia Pacific Institute for Theoretical Physics

Cosmomc with mock data

Post by joby pk » October 04 2016

Dear Antony,
I am using the June 2016 version of Cosmomc and running using mock data with default paramteres. The mock data was generated using the makePerfectForecastDataset.py code with the Cl's output from CAMB. For some reason the Monte Carlo does not start and I get the following error:

starting Monte-Carlo
Matrix_Cholesky: not positive definite 2
MpiStop: 23
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 23

I read up similar posts on Cosmocoffee where the issue was resolved by deleting the *.chk checkpoint files. However in my case, I have disabled checkpointing and hence there are no .chk files.

Any idea about what I could be doing wrong?

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Cosmomc with mock data

Post by Antony Lewis » October 04 2016

A problem with your propose matrix?

joby pk
Posts: 12
Joined: February 26 2016
Affiliation: Asia Pacific Institute for Theoretical Physics

Cosmomc with mock data

Post by joby pk » October 04 2016

Not sure... I am using defaults.

joby pk
Posts: 12
Joined: February 26 2016
Affiliation: Asia Pacific Institute for Theoretical Physics

Cosmomc with mock data

Post by joby pk » October 17 2016

So I gave a run of cosmomc with input dataset being the Cl's(TT) obtained from CMB sky simulation(based on Planck LCDM parameter values). I have used propose matrix= run.covmat generated from the previous run with the same parameters. At the end of the run, my R-1 value was 0.0011. When I use Getdist to analyse the chains, I notice that there is no proper "peak" in the Tau plot. Could you tell what is going wrong?
The range set for Tau in batch2/params_CMB_defaults.ini is: param[tau] = 0.09 0.01 0.8 0.01 0.005

I am attaching the Tau plot here.

/cosmo_files.php?file=joby_pk/tau.jpg

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Cosmomc with mock data

Post by Antony Lewis » October 17 2016

Perhaps your simulated data had tau value of 0? (so cut off by the 0.01 minimum). There doesn't have to be a peak away from the boundary.

Post Reply