CosmoCoffee Forum Index CosmoCoffee

 
 FAQFAQ   SearchSearch  MemberlistSmartFeed   MemberlistMemberlist    RegisterRegister 
   ProfileProfile   Log inLog in 
Arxiv New Filter | Bookmarks & clubs | Arxiv ref/author:

[CosmoMC + ModeChord] Suspiciously (very) tight posterior distributions
 
Post new topic   Reply to topic    CosmoCoffee Forum Index -> Computers and software
View previous topic :: View next topic  
Author Message
Mason Ng



Joined: 17 May 2017
Posts: 25
Affiliation: The University of Auckland

PostPosted: September 07 2017  Reply with quote

I am doing a quick test run with ModeChord where I am calculating the Bayesian evidences for the inflationary models with potentials m2φ2 and λφ4. When I run these and plot the chains, I get a very suspicious-looking triangle plot.



As seen above, the parameters related to inflation (ns, r, log(m2)*) have a really tight distribution. This is not seen in tau. I did not put this in the plot I attached, but log(As) does not have a tight distribution (i.e. something more sensible).

I was wondering if anyone has any idea about this? I attach a sample .ini file below. In one of the lines, note that I had "compute_tensors=T". This was set to F for /that/ triangle plot. It was only after I generated that triangle plot did I realize that compute_tensors should probably be set to T. I am also testing right now if putting a prior on Npivot would change things too. I'm running the code again, but I'm just wondering if anyone has any further comments on this, and/or if that was the last thing I missed.

* Note 1: I know that lphi4 uses λ for vpar1 and m2phi2 uses m2. I was just doing a quick plot.

Code:
#Use Planck 2015 likelihoods
DEFAULT(./batch2/plik_dx11dr2_HM_v18_TT.ini) #high-l likelihood
DEFAULT(./batch2/lowl.ini) #low-l likelihood; low TT WITHOUT Pol
DEFAULT(./batch2/lensing.ini) #for lensing
#DEFAULT(./batch2/BKPlanck.ini)

#other likelihoods
#DEFAULT(./batch2/BAO.ini)

#general settings
DEFAULT(./batch2/common.ini)

## INFLATION MODEL TO USE
#DEFAULT(Models/m2phi2/m2phi2.ini)
DEFAULT(Models/lphi4/lphi4.ini)

## Reheating settings:
## Instant reheating
DEFAULT(batch2/modecode_inst.ini)

## Defaults
DEFAULT(batch2/modecode_defaults.ini)
DEFAULT(batch2/modecode_adjustments.ini)

# Many model parameters can be 'un-physical'.
# We make use of cosmomc's logzero functionality, but this requires that it
# doesn't stop when it hits an error
stop_on_error=  F

#high for new runs
MPI_Max_R_ProposeUpdate = 30

propose_matrix = ./planck_covmats/base_TT_lensing_lowl_plik.covmat

#folder where files (chains, checkpoints, etc.) are stored
root_dir = ./chains/

#root name for the files produced
file_root = lphi4_tensors_cal

#action = 0 to MCMC, action = 1 to postprocess .data file
#action = 1 for important samples, action = 4 to quickly test likelihoods
#action = 5 for the PolyChord sampler
action = 5
DEFAULT(batch2/polychord.ini)
nlive = 500

num_threads = 8

#to get theory cl for test point
test_output_root = output_cl_root

start_at_bestfit = F
feedback = 1
use_fast_slow = T

checkpoint = T

#sampling_method = 7 is a new fast-slow scheme good for Planck
sampling_method = 7
dragging_steps = 3
propose_scale = 2

#Set >0 to make data files for importance sampling
#26 August - don't need this yet
indep_sample = 0

#small speedups for testing
get_sigma8 = F
#already F in batch2/modecode_adjustments.ini

#REMEMBER, to use lensing, turn OFF semi slow parameters.
######## Semi Slow parameters ON #############
# Turn on semi-slow parameters
#use_nonlinear_lensing = F
#block_semi_fast = T

# Set up the fraction of time to spend on each type of parameter
#nest_frac_slow      = 0.75
#nest_frac_semi_slow = 0.20
#nest_frac_semi_fast = 0.00
#nest_frac_fast      = 0.05


######## Semi Slow parameters OFF #############
# Remove the equivalent lines directly above, and uncomment
# the lines below in order to turn on non-linear lensing
# (and turn off semi slow parameters)

# # Turn off semi-slow parameters
use_nonlinear_lensing = T
block_semi_fast = F
#
# # Set up the fraction of time to spend on each parameter
nest_frac_slow      = 0.95
nest_frac_semi_slow = 0.00
nest_frac_semi_fast = 0.00
nest_frac_fast      = 0.05

## priors for physical parameters
H0_min = 20
H0_max = 100

#to vary parameters, set param[name] = center, min, max, start width, propose_width
#for fixed, can just fix fixed value

## ModeCode-modified prior ranges
#param[omegach2] = 0.12 0.08 0.16 0.001 0.002
#param[omegabh2] = 0.02225 0.02 0.025 0.0001 0.0002
#param[theta] = 1.0411 1.00 1.08 0.0004 0.0004
#param[tau] = 0.09 0.05 0.13 0.001 0.001

#neutrino_hierarchy = degenerate
#num_massive_neutrinos = 1
#param[mnu] = 0.06

#param[omegak] = 0
#param[w] = 0
#param[r] = 0
#param[nrun] = 0
#param[nrunrun] = 0

#MODIFIED P(K)
#inflationary potential parameters
#N_pivot
param[N_pivot] = 50
#vparams array
param[vpar1] = -13.4 -16.0 -10.0 0.1 0.1
param[vpar2] = 0
param[vpar3] = 0
param[vpar4] = 0
param[vpar5] = 0
param[vpar6] = 0
param[vpar7] = 0
param[vpar8] = 0
param[vpar9] = 0
param[vpar10] = 0
param[vpar11] = 0
param[vpar12] = 0
param[vpar13] = 0
param[vpar14] = 0
param[vpar15] = 0
param[vpar16] = 0
param[vpar17] = 0
param[vpar18] = 0
param[vpar19] = 0
param[vpar20] = 0
# END MODIFIED P(K)

param[omegabh2] = 0.02225 0.02 0.025 0.0001 0.0001
param[omegach2] = 0.12 0.08 0.16 0.001 0.0005
param[theta] = 1.0411 1.00 1.08 0.0004 0.0004
param[tau] = 0.09 0.01 0.20 0.001 0.001

param[logA] = 3.1 2.5 3.5 0.001 0.001
#log[10^10 A_s]
param[ns] = 0.96 0.8 1.2 0.004 0.002
inflation_consistency = T

param[omegak] = 0
param[w] = -1

#altered on 7 Sept
param[r] = 0.03 0 2 0.001 0.001
compute_tensors = T
#

param[nrun] = 0
param[nrunrun] = 0

neutrino_hierachy = degenerate
num_massive_neutrinos = 1
param[mnu] = 0.06
Back to top
View user's profile  
Mason Ng



Joined: 17 May 2017
Posts: 25
Affiliation: The University of Auckland

PostPosted: September 11 2017  Reply with quote

The cause is about the code somehow fixing Npivot somewhere - I am testing this at the moment I'll update on this, though I'm getting an error following on from the above. I'll put this on a new thread.
Back to top
View user's profile  
Display posts from previous:   
Post new topic   Reply to topic    CosmoCoffee Forum Index -> Computers and software All times are GMT + 5 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group. Sponsored by WordWeb online dictionary and dictionary software.