1,699 followers
@SLKoole Did this awhile back with a large combined data set (n=~2000). Split in half with model development in first half, replication of final models in hold-out. Calculated bias and abs. diff. in parameters across halves. Focused on r>.3 effects g…