# What kind of independent variables can I use for multiple regression?

I'm very new to statistics. My assignment requires me to use one statistical method taught during lesson so I only have a choice between multiple regression, logistics regression and MANOVA.

My research proposal is to look at the effects of violent video games and aggression. My lecturer suggested me to use multiple regression, so I can look at what factors predict aggression when playing video games.

After reading through several journal articles, I've come out with a few factors that may affect aggression, the types of violent video games played (collaborative or competitive), realism (how much the game relates to reality), family history and gender.

I could set aggression as my dependent variable but I'm not sure how my independent variables could work. Looking at the list of independent variables I came out with, I need to let my participants play a few kind of games (collaborative vs competitive and realism vs non-realism). Then when should I measure their aggression, after they played each game? Wouldn't that make it two dependent variables?

I'm really confused. If there's a way to do this with logistic regression or MANOVA please help me with it too! Thanks!

• Can you have all of your participants play all of the games?
– EdM
May 17, 2015 at 18:38
• I can. But the dependent variable is their aggression after they played each game. If I do so, then I would have to measure the dependent variables a few times? Unless I can use different dependent variable? May 17, 2015 at 18:41

I'm assuming here that what you seek to measure is the aggression measure after an individual plays a particular game, rather than some broader measure of overall aggressiveness.

For each of your combinations of games and participants, you will get an additional value of the dependent variable of aggression, but these measurements are (ideally) all samplings from the same single random variable of your aggression measure. That's the distinction that you might be missing. Make sure to review what you were taught about the nature of random variables.

Multiple regression of the type you envision seeks to find how the value of a dependent random variable is associated with each of the independent variables. To accomplish this task, you have to measure the value of the single dependent random variable under a variety of circumstances (combinations of independent variables), as you clearly recognize.

Since you can have each of the participants play each of the games, it will be straightforward to consider the identity of the participant as an independent variable, along with gender, family history, and the characteristics of the game. Your course ideally taught you how to take advantage of this type of experimental design.

Since this is apparently for a course assignment I'll stop the answer here. You will learn this best by doing it yourself. Unless this exercise is part of an exam during which further contact with the lecturer is prohibited, check in frequently with the lecturer to make sure you are not going too far astray. If it is an exam, make sure to talk with the lecturer when all is done to see what you did and didn't understand.

• Thanks a lot! I realized where I was lost. They didn't taught us about dependent random variable. None of the examples they gave had an independent variable used as between or within subjects. I've tried to do a Google search but I can't seem to find any articles on dependent random variable. Do you happen to have any links about it? May 17, 2015 at 19:44
• I'm able to consult my lecturer's help for this assignment. The issue is it's midnight here and I kind of need answers urgently because I don't have much time left to do this. If they didn't taught us about dependent random variable, perhaps it's not ideal to write this in my assignment. I've thought of another way to go about doing this, can you help me see if it's feasible? First, I would ensure both the collaborative and competitive violent video games have the same realism, so that's one extraneous variable removed. May 17, 2015 at 19:45
• Then, I'll randomly assigned half the participants to play collaborative violent video game and the other half to play competitive violent video game. This will be my first predictor. The rest of the predictors will be gender and family history. May 17, 2015 at 19:46
• After they've played the games, I'll measure their aggression level. In such a case, I don't have to worry about having different dependent variables. Is this method feasible? Thanks. May 17, 2015 at 19:46
• I don't see that you have to deal with "different dependent variables" in any case; you just are measuring values of the single "aggression" variable under different conditions. You can split the participants into 2 groups, but you will lose information from how the same person responds to the different types of games.
– EdM
May 17, 2015 at 20:00