Willman I vs. the Nmix

I don't feel like typing up each individual stat, as there are a lot for this galaxy. So I'll just copy and paste the relevant Nmix data from the .pe file (see previous posts to understand what the numbers mean):

1 173881
1.00000 -13.49319 6.00589

2 283751
0.56915 -17.32806 3.94248
0.43084 -8.08833 3.94228

3 198372
0.36418 -19.62466 3.61210
0.33942 -13.29231 3.86374
0.29639 -5.97242 3.56708

4 124760
0.24127 -21.75585 3.47407
0.29321 -15.86464 3.56630
0.24968 -10.76959 3.66620
0.21583 -4.12989 3.38790

What's interesting about this is that both the double and triple component fits are more probable than the single component fit. Of course, this is without taking metallicity into account, which could possibly remove the stars that are Milky Way contaminants.

But from looking at the histograms with the fits plotted, the two-component model looks pretty darn good. But that relies on my scaling of the curves by eye as I can't really figure out how to truly scale them.

I did the same thing for the Willman I data without the high metallicity contaminant stars and got 21% for one component, 26% for two, and 18% for three. So it's pretty likely that there are multiple populations in there.

TO DO:
1) Write a paragraph about Nmix--what type of statistical grouping test it is
2) Bootstrap the Willman I data--take a data set of 40/45 (40 if we don't include the high metallicity stars) of the stars with replacement, then Nmix that, record results
3) Find some way to program the above into IDL
4) ?????
5) PROFIT!