0
$\begingroup$

One of the assumptions for performing ordinal regression is linearity. How to test this for this specific type of regression? Do you have to use logit etc.? And how does this work in SPSS?

$\endgroup$
3
  • $\begingroup$ Do you mean "linearity" in terms of the relationship between the log-odds of the outcome and its predictors, or the proportional odds (or parallel regression) assumption, which can be assessed graphically or using dedicated test (e.g., Brant test)? $\endgroup$
    – chl
    Nov 21, 2020 at 7:26
  • $\begingroup$ The first one, relationship between log-odds of the outcome and its predictors. I assume the latter is tested using the spss output of the ordinal regression analysis by looking at the test of parallel lines outcome? $\endgroup$
    – Chris
    Nov 21, 2020 at 8:26
  • $\begingroup$ The Brant test is anti-conservative. See Peterson and Harrell (1990) and fharrell.com/post/rpo $\endgroup$ Sep 19, 2022 at 12:23

1 Answer 1

1
$\begingroup$

There's nothing special about ordinal regression models in this regard. You would assess linearity in the same manner as you would for a linear or binary logistic regression, comparing results for variables as they are vs. appropriate transformations of them and/or combinations of predictors. There's no automated way to formally test this in any of the SPSS regression procedures (or in those in other packages either, as far as I know).

$\endgroup$

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.