I have classroom network data where kids named their friends in the classroom (a directed network). I'd like to measure the closeness and betweenness centrality of each kid within their classroom. In some classrooms, there are kids who report friends but who themselves were unidentifiable, likely because their wrote their name wrong or slopily. Should I still incorporate information from these individuals?
My intution says yes. But I could see an arguement against it. The resulting measures when keeping vs dropping the nominations from these individuals are highlighly correlated (>.75) but I don't know which measures (i.e., those with vs. without these nominations) make the most sense to focus on.