Hierarchical regression in SPSS (how to enter interactions when you have a quadratic)

[tropie]

Hello there!

I want to confirm if I am entering my variables in the proper way in SPSS for multiple regression. I have four variables: 1) student gender (dummy); 2) student gender proportionality (continuous); 3) faculty gender (dummy); and 4) student gender proportionality SQUARED (continuous).

I am interested in investigating the main effects and the interactions between the variables. My understanding from my old stats classes and an extensive online search is that when doing hierarchical regression you first enter the control variables, then the predictor variables, then the interaction terms. However, what do you do when you have two AND three way interactions AND some of the interactions contain a quadratic term?

Below is my strategy to enter the variables into SPSS but I’m not certain. I’ve also looked at several academic journal articles but NONE of the articles that I’ve found have the quadratic term as part of an interaction (perhaps that’s a clue to something I may be misunderstanding?). Further, this situation is not referenced in any of my stats books or articles.

Strategy for entering variables into SPSS:

1. First level: control variables
2. Second level: main effects (e.g., student gender, student gender proportionality and faculty gender)
3. Third level: interactions between second level main effects plus quadratic term by itself? (student gender X student gender proportionality; student gender X faculty gender; student gender proportionality X faculty gender; student gender proportionality SQUARED)
4. Fourth level: three way interaction and two way interactions with quadratic term? (student gender X student gender proportionality X faculty gender; student gender X student gender proportionality SQUARED; faculty gender X student gender proportionality SQUARED)
5. Fifth level: three way interaction with quadratic term (student gender X faculty gender X student gender proportionality SQUARED)

Is this the correct way to do this? Any stats books / guides that might help me out? I appreciate your help with this and I'll be sure to post the solution once I find one! Thanks so much!

 

[noetsi]

Re: Hierarchical regression in SPSS (how to enter interactions when you have a quadra

Does the top half of this help..

https://books.google.com/books?id=P...ction effects hierarchical regression&f=false

 

[tropie]

Re: Hierarchical regression in SPSS (how to enter interactions when you have a quadra

Hi Noetsi,

Thanks so much for taking the time to respond. Yes, that confirms what I thought about how to enter variables but it doesn't speak to how to handle quadratic terms. Yikes. I did find one psych journal article that first entered all of the predictor variables, then the interactions between the predictor variables, then the quadratic term alone, then the quadratic term in an interaction. Phew! I think I'm going to follow that model given that it was recently published in a highly-regarded journal. Thanks so much!

 

[noetsi]

Re: Hierarchical regression in SPSS (how to enter interactions when you have a quadra

Your welcome. I don' t think there are any "rules" for this - none that are agreed on anyway. I doubt personally that there is any agreed on theory behind it. So what you do is all you can do, find an author who is well respected and cite them for your decision (of course it is even better if you know their logic for doing so - if there is any). I searching this topic, we had a similar question a few weeks ago, I never saw the logic for the decision described.

You can always email the author and ask them what the logic is (but that is brave because if they don't really have any it could be an embarassing conversation).