I have a model measuring Click through rates using 3 dummy variables.

  1. Placement location (PL1 vs. PL2)
  2. Ad type (Text vs. RM)
  3. Device type (Mob vs. Desk)

Additionally I want to measure the interaction effect of placement location x device type and placement location and ad type.

Therefore the following interaction dummies are made:

PL1xMob - PL1xDesk - PL2xMob - PL2xDesk (leaving one out)

PL1xText - PL1xRM - PL2xText - PL2xRM

Using SPSS and including all dummies in the linear regression. SPSS automatically kicks out the original dummies. Also because both interactions have Placement location as an interaction variable, SPSS trows out multiple interaction dummies.

Is there a way to measure the interaction effects of both interactions in one linear regression model? Or how should I construct the regression?


1 Answer 1


What SPSS is doing is correct. When you use dummy variables, one level will be the reference level and be removed. You compare the result for the other levels to that reference level. This is also true for interactions - one combination will be reference level


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.