Likelihood differences between Kelly 2007 and what we have: Only a normalization difference of sqrt(1+m^2). Variable swaps work out if we use the correct prior. The likelihood stays the same in our case and not in Kelly 2007, strangely.
Edge effects.
How to deal with data with arbitrary cuts.
Tests to do:
Generate the points with x belongs to [-5,5], m=0.5, c=0.0 and see how the MCMC performs in the different cases.
Gaussian mixture model may fail when dealing with power law selection function. Needs to be verified. Write down your way of thinking about the non-detections : I have this now... Involves some error function on x.
Preparing the FOF code to go to the Santa Barbara wiki.
Available code:
1) Calculate the overdensity as a function of halo mass.
2) Calculating corrections based upon the simulation resolution -> Decided against making this public as it may be misused.
August 25, 2011:
Line fitting: Think about non-detections.
August 26-29, 2011:
Distractions from dSph data.
August 30, 2011: Back to line fitting.
Think about:
Likelihood differences between Kelly 2007 and what we have: Only a normalization difference of sqrt(1+m^2). Variable swaps work out if we use the correct prior. The likelihood stays the same in our case and not in Kelly 2007, strangely.
Edge effects.
How to deal with data with arbitrary cuts.
Tests to do:
Generate the points with x belongs to [-5,5], m=0.5, c=0.0 and see how the MCMC performs in the different cases.
Gaussian mixture model may fail when dealing with power law selection function. Needs to be verified. Write down your way of thinking about the non-detections : I have this now... Involves some error function on x.