The empirical validations performed thus far would suggest … [object-oriented] metrics is close to being identified. However, none of these … strong size confounding effect and question the results … we perform a study on a large C++ … Our findings indicate that, before controlling for size, … After controlling for size, none of the [object-oriented] metrics we studied were associated with fault-proneness anymore … . [p. 642] researchers in this area … are urged to reexamine their data and determine … studies should always control for size.
Google-1
Google-2
Copyright clearance needed for quotation.