Error in CosmoMC regarding BAODR12 dataset

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
Alvaro Hu
Posts: 2
Joined: August 19 2016
Affiliation: University of Dallas

Error in CosmoMC regarding BAODR12 dataset

Post by Alvaro Hu » August 19 2016

When I run jobs using the BAO.ini in batch 2 with BAODR11 and test the likelihoods, I receive this:

loglike chi-sq
6.079 12.157 CMB: lensing = smica_g30_ftl_full_pp
581.392 1162.783 CMB: plik = plik_dx11dr2_HM_v18_TT
5249.288 10498.575 CMB: lowTEB = lowl_SMW_70_dx11d_2014_10_03_v5c_Ap
7.639 15.278 CMB: lowl = commander_rc2_v1.1_l2_29_B
5.439 10.878 Hubble: H073p0
65.299 130.597 SN: Union2.1
0.044 0.087 BAO: 6DF
0.463 0.927 BAO: MGS
1.428 2.856 BAO: DR11CMASS
0.618 1.236 BAO: DR11LOWZ

But when I run the BAO.ini in batch 3, with BAODR12, the more recent dataset, I receive this:

loglike chi-sq
6.079 12.157 CMB: lensing = smica_g30_ftl_full_pp
0.000 0.000 CMB: plik = plik_dx11dr2_HM_v18_TT
5249.288 10498.575 CMB: lowTEB = lowl_SMW_70_dx11d_2014_10_03_v5c_Ap
0.000 0.000 CMB: lowl = commander_rc2_v1.1_l2_29_B
0.000 0.000 Hubble: H073p0
0.000 0.000 SN: Union2.1
0.000 0.000 BAO: 6DF
0.463 0.927 BAO: MGS
0.110 0.220 BAO: WIGGLEZ = WiggleZ
0.000 0.000 BAO: DR12CMASS
0.459 0.917 BAO: DR12LOWZ

where a lot of the likelihoods are 0.000. The only thing I changed was going from BAODR11 to BAODR12. Also, when I run the chains for BAODR12, I receive an error saying that it could not start after 1000 tries, but it works for BAODR11.

Antony Lewis
Posts: 1364
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Error in CosmoMC regarding BAODR12 dataset

Post by Antony Lewis » August 25 2016

Looks like something more general messed up, e.g. try a clean build (and check plik tests return sensible likelihoods when run separately).

María Pía Piccirilli
Posts: 5
Joined: September 17 2016
Affiliation: Facultad de Ciencias Astronómicas y Geofísicas - Universidad de La Plata

Error in CosmoMC regarding BAODR12 dataset

Post by María Pía Piccirilli » September 17 2016

Hi everyone!
I'm having the same trouble with BAO datasets.
Plank data alone or together with BAODR11 runs perfectly, but when I try BAODR12 in batch3 I get the error message: "MCMC.f90: Couldn't start after 1000 tries - check starting ranges".
I've tried a clean build with the same results as before.
Here I copy the output of the likelihood test, having exactly the same problems as cited in the first post by Alvaro.
Any clue?
Thanks in advance!

plik TT + lowTEB:
loglike chi-sq
581.156 1162.311 CMB: plik = plik_dx11dr2_HM_v18_TT
5249.287 10498.574 CMB: lowTEB = lowl_SMW_70_dx11d_2014_10_03_v5c_Ap
Test likelihoods done, total logLike, chi-eq = 5830.568 11661.135

plikTT + lowTEB + BAO DR11:
loglike chi-sq
581.156 1162.311 CMB: plik = plik_dx11dr2_HM_v18_TT
5249.287 10498.574 CMB: lowTEB = lowl_SMW_70_dx11d_2014_10_03_v5c_Ap
0.044 0.088 BAO: 6DF
0.463 0.927 BAO: MGS
0.110 0.221 BAO: WIGGLEZ = WiggleZ
1.431 2.861 BAO: DR11CMASS
0.620 1.240 BAO: DR11LOWZ
Test likelihoods done, total logLike, chi-eq = 5833.236 11666.472

plikTT + lowTEB + BAO DR12:
loglike chi-sq
0.000 0.000 CMB: plik = plik_dx11dr2_HM_v18_TT
5249.287 10498.574 CMB: lowTEB = lowl_SMW_70_dx11d_2014_10_03_v5c_Ap
0.000 0.000 BAO: 6DF
0.000 0.000 BAO: MGS
0.110 0.221 BAO: WIGGLEZ = WiggleZ
0.000 0.000 BAO: DR12CMASS
0.000 0.000 BAO: DR12LOWZ
Test likelihoods done, parameter point rejected (logZero or outside prior)

Alvaro Hu
Posts: 2
Joined: August 19 2016
Affiliation: University of Dallas

Error in CosmoMC regarding BAODR12 dataset

Post by Alvaro Hu » September 18 2016

What I ended up doing was replacing the BAODR12.ini in the job .ini file to BAODR12_RSD.ini, which allowed it to run. Hopefully that works for you as well

Antony Lewis
Posts: 1364
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Error in CosmoMC regarding BAODR12 dataset

Post by Antony Lewis » September 19 2016

Check you have the latest version from github, cf.

https://github.com/cmbant/CosmoMC/issues/6

María Pía Piccirilli
Posts: 5
Joined: September 17 2016
Affiliation: Facultad de Ciencias Astronómicas y Geofísicas - Universidad de La Plata

Error in CosmoMC regarding BAODR12 dataset

Post by María Pía Piccirilli » September 21 2016

Thank you very much Alvaro and Antony for your reply!

I've tried both suggestions. Working with RSD files leads me to something strange with the use of memory, as for running just a likelihood test requires more than 60gb which leads to the job beeing killed by the cluster queue manager.

Replacing data/sdss_DR12CMASS_bao.dataset and adding data/sdss_DR12CMASS_bao_dilation_qpm.dat and data/sdss_DR12CMASS_bao_dilation_md-patchy.dat downloaded from https://github.com/cmbant/CosmoMC/pull/ ... 3b3c4a0e49 didn't work at first. I tried getting the new version of bao.f90 and recompiling, but I get:
bao.f90(27): error #6366: The shapes of the array expressions do not conform. [MEASUREMENT_TYPES]
character(LEN=Ini_Enumeration_Len), parameter :: measurement_types(8)

What am I missing?
Thanks!

Antony Lewis
Posts: 1364
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Error in CosmoMC regarding BAODR12 dataset

Post by Antony Lewis » September 21 2016

I would just do a new full clone of the latest git repository rather than trying to get individual files.

María Pía Piccirilli
Posts: 5
Joined: September 17 2016
Affiliation: Facultad de Ciencias Astronómicas y Geofísicas - Universidad de La Plata

Error in CosmoMC regarding BAODR12 dataset

Post by María Pía Piccirilli » September 23 2016

Thank you very much, Antony!
A full clone worked perfectly!

Post Reply