Overlapping Independent Variables When two independent variables measure similar characteristics and are highly predictive of the dependent variable, stepwise regression is not the best choice. For example, researchers might explore whether hours spent playing violent video games and hours spent watching violent television shows— independent variables that are strongly related to each other—predict aggression levels in children. Stepwise regression will likely show that one of these variables—say, video game playing—is a strong predictor of aggression; the second variable, watching violent TV shows, won’t explain any additional variability because it overlaps so much with the first. The first variable, video game playing, gets credit, so to speak, for all of the variance it contributes itself as well as all of the variance it shares with watching violent TV shows. The regression will falsely indicate that only violent video game playing predicts aggression.