Given the potential for confounding factors - some of which are listed in(b) - finding a negative correlation would not be strong evidence that smaller classsizes actually lead to better performance. Some way of controlling for the confound-ing factors is needed, and this is the subject of multiple regression analysis