I apologize for the vagueness in this example, I thought it'd be clearer using the character values instead of saying I need 33=1, 39=2, and so forth since the values are not continuous 33, 34, 35, etc.). In the study I'm trying to replicate, they used dummy variables for the product type (though I'm unsure why, but I assume there's a reason that would affect my results?).
Apologizing is nice, but providing a clear problem statement is even better. Anyway, changing 33 to 1, and 39 to 2, and so on, is pointless, and gains you nothing. Now perhaps you have the situation where BOTH 33 and 67 get changed to 1, in which case such a transformation is useful and not pointless; but you didn't say that. So I will stick with my pointless statement.
In the study I'm trying to replicate, they used dummy variables for the product type (though I'm unsure why, but I assume there's a reason that would affect my results?). The product type is meant to be a control variable, but from what I understand it functions just like any other independent variable. I can run a regular OLS regression with the variable as-is, without the dummy variables, but I'm not getting the correct results - they are actually way off, not just a little bit. So I was trying to figure out another way.
Let's forget this whole dummy variable thing for now. It is not relevant to describing the problem clearly, and getting good results. Furthermore, as I stated earlier, many SAS procedures, including PROC GLM (which is the PROC I think you should use) does not need dummy variables to be created beforehand. GLM will create the dummy variables for you, so you don't have to.
If you are not getting correct results, there are many possible reasons for this, and so you can't simply conclude that you need to add dummy variables.
Let me ask you this ... are you using the exact same data as the study you are trying to replicate? Or are you using your own data but using the design of the study you are trying to replicate? This isn't clear either.
... View more