New Correlation Results
Today we ran the results from the new correlations for sheriff hours. We have found that sheriff hours do not correlate to degree, closeness, or betweenness when looking at an overall spearman correlation. The lack of correlation is even MORE true if we only look at non-zero sheriff hours. We also looked for a correlation with the mann-whitney-wilcoxon test, where we looked at the median of degree/closness/betweenness with sheriff hours and without sheriff hours.
We found that the median degree with sheriff_hrs is 248 and the median degree w/o sheriff_hrs is 10.0 The wilcoxon test says that the p-value is 0.0. From this information we can conclude that for sheriff hours and degree, there is no correlation: as sheriff_hours increases, we cannot say that degree will increase at some rate. However, we were able to learn that non-sheriffs do have lower degrees than sheriffs, though because there are so many outliers, there is no correlation. (There are a lot of non-sheriffs with very large degrees, and a lot of sheriffs with low degrees.)
We found similar results with sheriff hours and closeness. They are not correlated and the wilcoxen test medians with sheriff_hrs was 0.4688 while the median w/o sheriff_hrs: 0.3822. It also had a p-value of 0.0. This tells us that sheriff hours and closeness has no correlation relationship.
Our results for sheriff hours and betweenness were quite different. Although we saw in spearman that the correlation was the weakest of the three, wilcoxen told us that the median betweenness with sheriff_hrs is: 0.0013, while the median betweenness w/o sheriff_hrs is: 2.355e-06 with a p-value of 3.086917507911853e-275. This means that when a user becomes a sheriff, they become more between in the network. However this is not a correlation: as sheriff hours increases, the betweenness does not necessarily increase. But, if a developer is a sheriff, they are likely to be more between.
All this information is very helpful to our research. Output of the results are shown below.
----------------------------------------------------------------
------------SOCIAL NETWORK ANALYSIS: sheriff_hours--------------
----------------------------------------------------------------
----------------------------------------------------------------
------------Spearman on deg/closeness/betweenness---------------
----------------------------------------------------------------
closeness vs betweenness: 0.7099071135048419
closeness vs degree: 0.8732619561282218
betweenness vs degree: 0.8680034491199055
----------------------------------------------------------------
------Spearman on sheriff_hrs VS deg/closeness/betweenness------
----------------------------------------------------------------
degree vs sheriff hours: 0.47810605797734707
closeness vs sheriff hours: 0.4884067056335153
betweenness vs sheriff hours: 0.44583852037976784
----------------------------------------------------------------
-Wilcoxon p-values on sheriff_hrs VS deg/closeness/betweenness--
----------------------------------------------------------------
degree vs sheriff hours p-value: 0.0
median with sheriff_hrs: 248
median w/o sheriff_hrs: 10.0
closeness vs sheriff hours p-value: 0.0
median with sheriff_hrs: 0.4688
median w/o sheriff_hrs: 0.3822
betweeness vs sheriff hours p-value: 3.086917507911853e-275
median with sheriff_hrs: 0.00132714
median w/o sheriff_hrs: 2.355e-06
****************************************************************
----------------------------------------------------------------
-----What if we ONLY look at non-zero sheriff hour values?------
---------------------------Spearman-----------------------------
degree vs sheriff hours: 0.19458279332037892
closeness vs sheriff hours: 0.192259506330964
betweenness vs sheriff hours: 0.17563196331578557 ```