Multiple Regression

David C. Howell

This link is only barely started, and more will follow.

• If you want a whole set of variables with a specified set of correlations, check out the response that David Nichols wrote in response to such an inquiry.

• An complete implementation of David Nichols suggestion from the previous item can be found at CorrGen2.html.

• For an example using multiple regression, look at a discussion of the relationship between religious belief and optimism. This is based on a study by Sethi and Seligman, and is fairly extensive.

• An example built on a study by Geller, Johnston, and Madsen (1997) contains a  nice compilation of bits and pieces of multiple regression. It looks at hierarchical regression, tolerance, interaction terms, and Baron and Kenny's (1986) distinction between mediator and moderator effects.

• A topic of considerable interest these days is the analysis of interaction terms in multiple regression. If the relationship between stress and illness itself depends on the level of social support that the patient has, then the correct model would be of the form Illness = b1*Stress + b2*SocSupp + b3*(Stress*SocSupp) + b0. A wonderful program to illustrate this graphically is a freeware program called ITLASS. It can be found by going to http://www.simstat.com. Check it out.

• I have added a new page dealing with setting confidence limits on squared semi-partial correlations. This page adopts an asymptotic procedure developed by Alf and Graf (1999).

Return to Dave Howell's Statistical Home Page  

Last revised: 10/26/01