10.5 Take-Home Points

• In a multiple regression model, a regression coefficient represents the (predictive) effect of a variable while controlling for the effects of all other predictors. It is called a partial effect: It predicts variance of the dependent variable that cannot be predicted by the other predictors.

• If a new predictor is added to a regression model, the regression coefficient of an old predictor changes if the new predictor is correlated with both the old predictor and the dependent variable. If the old predictor’s effect becomes stronger, the new predictor was a suppressor. If it becomes weaker (the old effect was—partially—spurious) or changes direction (sign), the new predictor was a reinforcer.

• Random assignment of participants to experimental treatments (the independent variable/predictor in an experiment) is meant to create (near) zero correlations between the predictor and any other variable not included in the experiment. As a result, we expect that there are no confounders.