Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2016 + GetDist plot issues

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Luke Hart
Posts: 70
Joined: July 13 2015
Affiliation: University of Manchester

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Luke Hart » April 17 2017

Hi Antony, everyone,

First of all - thank you very much for the help received so far when trying to utilise CosmoMC. It has been of a great help, especially from Antony and Sunny.

So I have resolved the problem from re: convergence diagnostics posted in this forum last week (or the week before!) -- problem was that I was fixing theta when trying to marginalise over H0, effectively removing a degree of freedom from my problem when I really wanted to reverse the roles of theta nad H0.

My first question is: how does one do this? How do you configure CosmoMC for marginalisation over H0 with derived theta as opposed to the other way round?


Second question: in GetDist (the GUI is awesome by the way) I'm getting some overlapping axes which make the plots look very untidy, particularly with [tex]\Omega_b h^2[/tex] and [tex]\Omega_c h^2[/tex]
How do you modify the scripts mentioned in GetDistGUI to account for this, either multiply the axes by 10 and divide the labels by 10 or to make the fonjts smaller? I'm a little bit lost on how to tweak this successfully.

Cheers in advance :)

Image

Jason Dossett
Posts: 97
Joined: March 19 2010
Affiliation: The University of Texas at Dallas
Contact:

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Jason Dossett » April 17 2017

There are quite a few steps in re-configuring some of the source code to vary H0, I don't quite have the time to type them all out in detail, but I do have the modified source code. You can get it here:

http://www.utdallas.edu/~jdossett/files ... _h0.tar.gz

The tarball contains modified versions of all the files that need to be modified (and a few new ini/paramnames files) in order to vary H0.

Unfortunately, I don't remember the exact version of CosmoMC this came from, (it is a 2015 or 2016 version though) so you may need to make some modifications.

I do want to make sure though that you are sure you want to do this. I made this source code for looking at some MPK observations without using the Planck observations. I strongly suggest if you are using CMB observations you use the CosmoMC code as is. The theta parameterization was chosen because it resolves some degeneracies in the parameter space when using CMB observations.

Luke Hart
Posts: 70
Joined: July 13 2015
Affiliation: University of Manchester

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Luke Hart » April 17 2017

Thank you Jason,

I probably won't go into this, I was more curious than anything else. Since when we fixed theta we got some seriously strong constraints but it's obvious now that in standard PLANCK runs the H0 derived parameter isn't fixed, whereas we were effectively fixing this and tightening our analysis.

I feel that H0 variations would be more sensible, as you say, for something else like constraining with BAO observations.
Is there any chance you have any clue on how to change tick labels as I've mentioned in my second question (nonetheless, thank you for helping with the first :) )

Jason Dossett
Posts: 97
Joined: March 19 2010
Affiliation: The University of Texas at Dallas
Contact:

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Jason Dossett » April 17 2017

Unfortunately, I have not played around much with the plotting routines so can't really be of help there.

Sunny Vagnozzi
Posts: 55
Joined: August 15 2016
Affiliation: Kavli Institute for Cosmology (KICC), University of Cambridge

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Sunny Vagnozzi » April 18 2017

Hi Luke! I have never used the GetDist GUI, only the "regular" GetDist. So, if you're willing to switch to that, I might be able to help ;-)

Basically anything regarding plotting with the scripts supplied by getdist (which typically are in results/) you have to go and modify by hand in python/getdist/plots.py. To understand where you have to modify, you would have to into one of the scripts GetDist provides you. Let me be more specific: it seems to me you're doing 2D plots so the GetDist script which would do it for you would be something like root_2D.py, which you would then run from the main CosmoMC folder as python results/root_2D.py of course. If I open one of my 2D.py scripts I see the following line

Code: Select all

g.plots_2d(roots,param_pairs=pairs)
which means that in python/getdist/plots.py I have to go and look for the plots_2d routine which in my case is on line 1616, and modify that in order to modify ticking, labelling, etc.

Hope this helps in case you decided to not use the GUI :)

Cheers,
Sunny

Antony Lewis
Posts: 1936
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Convergence diagnostics in PLANCK runs of CosmoMC June 2

Post by Antony Lewis » April 18 2017

You should almost never have to modify plots.py (unless there is a bug).

Much better to adjust any spacing etc after making the default plot (or most things can be done using the function arguments/or the plots settings object to set preferences).

Is the specific spacing issue using matplotlib 2? Changing the subplot size/label font size arguments may help.

Post Reply