Modify F behaviour #66
Description
Copying comments from SLACK:
think the data has gotten good enough
that it is time to abandon F in zdm
That said, I don't quite know what to replace it with
Oh, I do now (as I type):
A binned log-normal PDF with mean = <DM_cosmic> and variance the free parameter
How many bins? I don't know
What type of bins, e.g. in z, (1+z) , log z ? I don't know
Clancy:
@Prof X
@Joscha
Fully agree. But I think we should aim to implement multiple methods here. As in, we add a new parameter called "F_METHOD", with something like:
F_METHOD 0: current implementation
F_METHOD 1: do binning in z or 1+z (maybe need another parameter to govern that
F_METHOD 2: F(z) predicted by simulation X
F_METHOD 3: F(z) predicted by simulation Y
etc. Then, publish a paper simply asking which F method currently fits the data best.
NOTE: If we do this, I think we have to re-examine DM_host. It's a tad insane that DM_host is currently constant with z in the rest frame of the host, ie. penalised by (1+z)^-1 in our frame. My two suggested paths forward for this are to either go for a generically binned method, and to actually fit for \sigma_DM (z), i.e. combine host and cosmic into one; OR to take both F(z) and host parameters from a simulation. In that sense, we're really at the level of asking "is Simulation X better than sim Y". I really don't want to go down that path, but in some ways in's the most self-consistent, rather than mixing things up.