Dependent variables (ordinal): credit rating 1970 (cr70) and credit rating 1980 (cr80).

Here is what I want to do:

Regress cr80-cr70 = independent vars.

How could this be done and how could you interpret it!?

If the dependent variable is continuous it would be simple. But can you make a new var from the difference of two ordinal vars, and have that be the dependent var?

  • 1
    $\begingroup$ +1 It's a good question. Note that it is pertinent even to the case of "continuous" dependent variables: in many situations it is not the case that numerical differences have the same meaning or interpretation regardless of the levels of the original values. $\endgroup$
    – whuber
    Oct 11, 2013 at 18:03

1 Answer 1


I would be very suspicious of an analysis constructed this way. When you say that the credit rating is an ordinal variable, you are saying that the difference between two ratings at a given time is not a meaningful quantity; all that is relevant is that one is higher than the other. If that is so, then the difference between two ratings at two different times is equally meaningless.

A better way to do this would be to include the earlier credit rating as an additional predictor in an ordinal regression for the later rating.


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.