Is some research just too different to robustly average?