Dan's Segment 31

From Computational Statistics (CSE383M and CS395T)
Jump to navigation Jump to search

To Calculate / Class Activity

1. This was done with Kkumar but I don't think she has posted it yet, hopefully that gets done soon.

The basic idea behind the first part is that adjacent points should be more or less the same, and thus any difference between the two is due to the standard deviation. So if you take the difference between adjacent points and divide by two you have an estimate for the standard deviation. Taking the mean over all such pairs is a decent estimation for the standard deviation of the entire sample.

As far as the actual fitting goes, we first tried to get a GMM to work, but had some trouble finding the right syntax in python. We eventually switched to a polynomial model which had a lot more success. As expected, adding additional terms always results in a better fit. Without the actual code I don't know how many terms ended up being helpful according to AIC or BIC.