Regression Discussion > How to Create Dummy Variables for Multiple Regression
Hi Cry Baby!
Thanks for your question! Do you mean that you have a single variable that measures internet use, where respondents choose between never, rarely, occasionally, and often?
If so, are you saying you want to know whether respondents that fall in each of these categories are of different weight? I just want to make sure I understand the question prior to offering a response....
Jeremy Taylor
Yes, I have independent variable 1 which measures Google use on a scale, independent variable 2 which measures Google Scholar use on a scale, and the dependent variable which measures the number of articles the user has had published within the past 5 years (continuous variable).
Cry Baby
What happened to the role of weight in your model? In your initial question, it sounded as though that was a key variable in your analysis. I'm a bit confused.
Also, I would be cautious about including an indicator of "Google use" and "Google scholar use" in the same model, as they are likely to cause multicollinearity problems.
Jeremy Taylor
Weight is lbs. is a typo. I don't know where that came from.
Cry Baby
I see. So with both "Google Use" and "Scholar Use" both in the model, are you looking to control for one of them, or are they both target predictors of publications?
Jeremy Taylor
I want to throw them both in the model and see what happens.
Cry Baby
I see, I think I understand. So both use of "google" and use of "scholar" are on the "1-Never, 2-Rarely, 3-Occasionally, 4-Often" scale you mentioned previously, right? If that is the case, you could try including them in the model as numerical values to start (since they do have a numerically meaningful order to them; i.e. they aren't purely categorical, they are more ordinal). However, technically speaking they aren't actually continuous, so you could also justify dummy coding, in which case you'll want to pick a reference category and make K-1 dummy codes (with K equallying the number of levels of your variable [4 in this case]). In other words, you'd create three dummy codes for each.
For example, the dummy codes for scholar might be:
scholar.rarely
scholar.occas
scholar.often
...with "never" being your reference group. The syntax to accomplish that in SPSS would be (assuming your scholar variable is called "scholar"):
Do If (scholar=2).
recode scholar (2=1)(SYSMIS=SYSMIS)(ELSE=0) into scholar.rarely.
else if (scholar =3).
recode scholar (3=1)(SYSMIS=SYSMIS)(ELSE=0) into scholar.occas.
else if (scholar=4).
recode scholar (4=1)(SYSMIS=SYSMIS)(ELSE=0) into scholar.often.
end if.
EXECUTE.
Alternatively, you could try using the General Linear Model (Univariate) dialogue menus and use the "contrasts" options to have SPSS automatically create the dummy codes. To do the kind of recodes that I did above, you'd select the "Simple" and "First" options for both IVs. I hope this helps!
Jeremy Taylor
I'm looking at frequency of Internet use on a scale of 1-Never, 2-Rarely, 3-Occasionally, 4-Often and weight in lbs. How do I create dummy variables for Internet use in SPSS to run multiple regression?