0
$\begingroup$

I used SPSS to create dummy variables for three different shops. Shop 1 was coded 0, shop 2 was coded 1 and shop 3 was coded 2. In my output I see that the slope coefficients (B) for shop 2 and 3 are almost identical, with almost identical CI, t statistics and p values. What does that mean, and is it necessary to separate the dummy variables when slope coefficient are so similar? And what could I have done instead if not separate them in dummies?

$\endgroup$
  • 1
    $\begingroup$ You should only have two dummy variables for the three shops. $\endgroup$ – gueststata May 22 '14 at 14:52
5
$\begingroup$

Ideally you should dummy code it as two variables. Then the third variable gets accounted for by the intercept of the model. Assuming these two variables are the only variables in the model,the coefficients you obtain are the difference between each shop with the non-dummy-coded shop along with the overall average. So its the incremental value when one goes from shop3 to shop 1 or shop3 to shop2 (assuming you didnt dummy code shop3)

You might want to look into contrasts coding to know more about using dummy variables in a regression model http://en.wikipedia.org/wiki/Contrast_(statistics) . Do check the external links provided in the wiki page that provides a really good explantion on contrasts.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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