Page 1 of 1

CosmoMC warning: mismatch in integrated times (CAMB: CalcScalarSources)

Posted: November 18 2020
by srijita sinha
Hi,

In one of my CosmoMC, it is giving the warning message:

Code: Select all

WARNING: mismatch in integrated times (CAMB: CalcScalarSources)
but there is no error message in the chains log file. This error is staring after

Code: Select all

 Current convergence R-1 =   0.2225066      chain steps =        6010
I could not understand anything from cmbmain.f90 file.

In another, there is no warning or error message from CosmoMC, but the chains-2 and 3 log files are showing:

Code: Select all

CAMB called 19500 times; 3 errors
.
This error is staring after around 16600 times of CAMB.

I am using only Planck likelihood and both of the CosmoMCs are for a dynamical DE model with CPL parameterization. I cannot understand what is going wrong in either of them.

Please help.
Srijita

Re: CosmoMC warning: mismatch in integrated times (CAMB: CalcScalarSources)

Posted: November 18 2020
by Antony Lewis
It's probably harmless as the error is triggered for relatively low errors - likely caused by the non-LCDM background requiring tighter timesteps to do the integration very accurately. You can test and fine-tune CAMB's accuracy parameters for your model before running CosmoMC if you want to be sure.

Re: CosmoMC warning: mismatch in integrated times (CAMB: CalcScalarSources)

Posted: November 18 2020
by srijita sinha
Thank you for your reply.

Re: CosmoMC warning: mismatch in integrated times (CAMB: CalcScalarSources)

Posted: November 20 2020
by srijita sinha
Hi,

There are certain parameter values for which the following error is being generated:

Code: Select all

-0.996198831932876       1.071091424913879E-002
CAMB returned error HMCode INTEGRATE, Integration timed out
where the 2 values are for [math] and [math] of the CPL parametrization. These values do not give the best results when checked in CAMB. I made the CosmoMC to not stop at those errors.

These points are not rejected from the MCMC chain. How to make these points rejected? Or their presence in the MCMC chain is not a problem? These points are resulting in error msg in the log file:

Code: Select all

CAMB called 10600 times; 60 errors
etc.

What to do?

As discussed in viewtopic.php?f=11&t=94&p=6982&hilit=savvas#p6982, I need to add

Code: Select all

use Errors
in Calculator_CAMB.f90? But I cannot find any module named

Code: Select all

Errors.
What am I missing?

Is there any thumb rule of setting the upper and lower limits of the parameters, if I have the best-fit values?

Sorry if these have already been discussed. Any help would be great!

Srijita