I expect predictor A to negatively predict my dependent variable, and predictor B to positively predict the dependent variable. Can I include both predictors in a (linear) multiple regression model even though the variables are associated to the dependend variable in different "directions"?
Thank you in advance and apologies for my English.
Yes. If these relationships with the dependent variable are true, then your coefficient for A will just be negative and the coefficient for B will be positive. You can use the regression to understand the magnitude of the effect A and B have on the dependent variable.
The only real question here is if A & B are closely correlated to each other (positively or negatively matters not). If they are, then you probably only need one of them.
As long as A & B are relatively independent of each other (i.e. not closely correlated), then yes they will both do well in your multiple regression model.
Including two correlated dependent variables in a multiple regression model won't generally impact its ability to predict outcomes, but it can make interpretation of the correlation coefficients quite difficult.
