I'm looking to see if whether social network properties (such as different measures of centrality) predict a binary choice. The first part of the question is, what is the best method to do this? I have seen papers that simply extract network scores (i.e betweenness, reach etc) for individuals and put them into a regression model to see if they predict the outcome. However, I've been told that this approach is incorrect, because you need to run randomisations/permutations when predicting signifiance from network data. I've been told to look up MQAP, so if anyone has any suggestions/advice about that, that would be excellent.
The second part of the question is regarding missing data. The social networks are based on children in classrooms, and as is a common problem in this area, I averaged about 45% parental consent. The network data was taken from social cognitive mapping, where participants are asked to name which other groups/pairs of children they know are friends (in this way triangulating the network data). The problem is then, that I have networks where I have no attribute/outcome data for often over half of the individuals in the nwetwork. Given that I'm interested in whether network properties predicted the binary choice in the participants that took part, is there a way to get around this 'missing data'?
Sorry for such a long question, any help/tips would be greatly appreciated.