Kernel weight per spike: what contributes to it at the individual QTL level?