Another reason (in addition to those already mentioned) why these numbers are not easy to interpret: small routines [..] had 23% more errors per LOC If the

Another reason (in addition to those already mentioned) why these numbers are not easy to interpret: small routines [..] had 23% more errors per LOC If the smaller routines had, on average, 100/1.23 = 81% or fewer of the lines of the larger routines, then they still had fewer errors per routine.