search

HWRix / Haloprofile

Text and Figures to go along with the Xue et al 2015 analysis of the Galactic halo profile
0 stars 0 forks source link

Are all the fits equally good? #8

Open HWRix opened 9 years ago

HWRix commented 9 years ago

screen shot 2015-01-13 at 13 53 45

If I look at the two numbers in the blue box, I don't understand them: the triple power law profile has EXACTLY the same exponents in its 2 outer slopes. --> It is effectively a singly-broken power law. Why is its likelihood better by 35 (!) than the best fit singly broken law??

One reason could be the normalization. take e.g. identical r_break=30, alpha1=2.7, alpha2=4.2, compared to r_break1=30, alpha1=2.7, alpha2=4.2 + r_break2=50, alpha3=4.2. Those two cases should give you IDENTICAL likelihoods. Right? Checke whether they do. If not, check whether they give you the same normalization constant...

xuexx commented 9 years ago

You are right. It is the problem of normalisation. When I calculate triple-power-laws, I used 48X48X48 points in MrXDMXFEH, but previously, I always used 20X48X20.

HWRix commented 9 years ago

Good. See what happens if you use the same grid for the normalization integral in all cases, and report what comes out