During the last week I finished running the last of the correlation and regression tests, and we determined which tests made the most sense to include in the write-up. We’ve come up with a layout for our write-up, and starting last week and into this week, I’ll be putting together that write-up and all of the figures. The main tests we’re including are correlations (across the month), and a sample correlation across May to see if MN size across the MN affects read counts, the AOV test, regressions and a new test included last week — the KS test. This test compares the regression coefficient results of the after timeframe of the 6 MN with the overall timeframe of our two control nodes. This test is a non-parametric test that can compare samples of two different sizes to see if the two underlying one-dimensional probability distributions are similar.
This test, the AOV test, and the May correlation test were not significant, and only a few correlation tests were significant. We have hypotheses for why our results are not significant – mainly not enough data – which we will talk about in the results. And we’ll also add a component of what kinds of metrics were most useful for future studies in this field.