Today I decided to use Nmix on the velocity data for Segue 1 (1 million sweeps, burn-in of 1 million), giving me the third set of data in the Segue Folder.
This gave me a 68.74% probability of one component with a mean of 207.85 and standard deviation of 8.29 (I still need to figure out how to find the errors in the mean).
But there is a 17.38% probability of two components, the first composing 46.6% of the stars with a mean of 201.29 and sd of 7.82, and the second has 53.4% of the stars with a mean of 212.58 and sd of 7.54.
I made a graph of the histogram with these Gaussians plotted on it, but am currently failing at getting it onto the internet.
I then Nmix'd the Hercules data:
There is a 48.78% probability of one component with a mean of 45.35 and sd of 6.33.
There is also a 20.36% chance of two components, the first with 51.0% of the stars with a mean of 40.94 and sd of 5.46, the second with 49.0% of the stars with a mean of 50.20 and sd of 6.00.
There is still a 10.55% chance of three components, although it doesn't really work based on the histogram.
I think the high percentage of multiple groups is just because there are only 30 stars in the file and there isn't really that much to group.
But I still made a bunch of sweet and COLORFUL graphs that show the fits on the histograms. They're pretty cool.
I'll even throw in a To-Do List because they seem to be the newest rage among my fellow researchers.
1) Nmix Willman I and Bootes II and make cool plots.
2) Figure out how to find the error in the means given by Nmix.
3) Actually read more of the stat books I got from the library.
4) Discover why Coke Zero is so much tastier than Diet Coke, despite having zero calories. It's a mystery!