I have a data set representing interpersonal relations between members of a certain group. I calculated two sets of centrality measures for this population. To calculate values for set A, I used data for all 77 participants (nodes) of my study. For set B, I removed 4 nodes (they have different status than the remaining 73) and re-calculated all measures. Calculating measures for set A and for set B is independent (as far as I can tell), but measures within each set are dependent.
I am trying to determine whether excluding those 4 participants with different status will make a difference for my data. When I plot histograms for both distributions they look very much alike, but I would like to be able to use a significance test to quantify this difference rather than "eyeballing". Here are histograms showing the measure distribution without (left) and with (right) those 4 additional nodes:
Here is what I thought of so far:
- I wanted to use the Wilcoxon signed rank test, but the distribution of the differences between the two groups is not symmetrical.
- I was thinking about the sign test, but observations for each participant are not independent, i.e., measure for one person affects measure for another person.
What test would be appropriate in my scenario?