My question is about using QAP in situations where some areas of the relational matrix cannot logically take “1”, by design of the study population. Suppose you ask to employees of a company with four very distant locations around the globe something like “who do you go to lunch with?” – and ties across locations are just impossible. Only some ties are technically possible (those within location): is there anything special we should do when performing QAP regression in that case?
I am asking because, as I understand, the procedure would possibly permute members into groups of people who could never see each other in reality, which sounds counter-intuitive.
I am interested in any reference or other procedures
Thank you !
Grenoble Ecole de Management