Dear All,

My student Tom Richardson and I are doing work here at KCL modelling 
Dwarf Spheroidal Galaxies, critical systems for understanding the nature 
of dark matter. We need to make our MCMC approach more efficient, and we 
thought that this would be an ideal opportunity to try and see if anyone 
in London with experience of these things wants to collaborate with us - 
the kind of thing which is a goal of the London Institute of Cosmology.

The physics is the following:- We have found a new way to measure the 
dark matter in the cores of dwarf Spheroidal galaxies. Understanding the 
density of dark matter in the cores of these galaxies can in principle 
tell us how many gamma rays we expect to see if the dark matter is 
composed of WIMPS, whether the dark matter is cold or warm and whether 
the dark matter is self interacting. We have extended the normal Jeans 
analysis to include the fourth moment (i.e. kurtosis) of the stellar 
velocity distributions and have shown that this can reduce the 
uncertainty in the density profile. http://arxiv.org/abs/1305.0670 We 
have developed a likelihood function that jointly fits the density and 
anisotropy parameters from the Jeans analysis to variance and kurtosis 
measurements of line-of-sight velocity data.

With a hand-written random walk MCMC code written in Python we have 
found it particularly challenging to automate a proposal density that is 
dynamically learned and have wasted time trying to use trial and error 
to find suitable covariance matrices for fast convergence of the MCMC 
chains. In other cases we have simply accepted inefficient proposal 
densities and needed very large chains to satisfy convergence criteria. 
Whilst this has been manageable for the recent work referenced above we 
would like to expand to larger parameter spaces in the future. Plus we 
want to take part in a data challenge later in the summer and at present 
our chains are simply converging too slowly.

What we would like is to place our likelihood analysis inside something 
clever like CosmoMC (which neither of us know how to use) to generate 
posterior distributions for these parameters and see how it affects 
indirect detection prospects and things like whether dark matter is warm 
or self interacting. We are therefore looking for a collaborator with an 
interest in the nature of dark matter and a good working knowledge of 
CosmoMC or some other integrated MCMC programme.  We could have done 
this ourselves but it would take longer and we thought this was a good 
opportunity to both widen our net of collaborators and, after all, one 
of the stated goals of the London Institute of Cosmology is to lead to 
new collaborations.

Your commitment would be to help our chains run faster and in return you 
would of course become a co-author. If you wanted to take part in the 
data challenge later in the summer that would be great but any help 
would be very welcome.  If you are at all interested, please contact us 
for more details

Sincere regards,

Malcolm Fairbairn and Tom Richardson