2
$\begingroup$

I am doing a multinomial regression and trying to interpret the results: In the basic model there is only one binary predictor variable (0 = high risk scenario, 1 = low risk scenario), the dependent variable has 3 categories (strategy 1,2 or 3).

The output shows that the model is significant and most of the logits as well. However, I am wondering how to interpret the (significant) logits of the intercepts. This is how I am interpreting it: "The likelihood to choose strategy 1 over 2 in a risk-free scenario (the intercept?) is higher/lower, keeping the rest constant." It this the correct way to interpret it?

Thanks in advance.

$\endgroup$

1 Answer 1

2
$\begingroup$

In a multinomial logistic regression with 3 levels of the DV there ought to be two intercepts. How exactly these are defined depends on which is the reference level. These will be the value of the logit when the independent variables are 0, in your case, when risk is high.

I wrote a presentation on multinomial and ordinal logistic regression; it somewhat concentrated on SAS, but some may be useful even if you are using another package.

$\endgroup$
4
  • $\begingroup$ Thanks for the file. Nonetheless now I am a bit confused: Then the coefficient of the intercept and the coefficient of the independent variable (if 0) should be same, since they would both mean the same. But this is not the case in the output. In fact, one is positive and the other one negative. Reference level is 1 (strategy 1) by the way. $\endgroup$
    – user18399
    Jan 11, 2013 at 17:42
  • 1
    $\begingroup$ The coefficients of the IV are not for different levels of the IV (since there are only two) they are for different levels of the DV. Just like the different intercepts are for different levels of the DV. Each level of the DV (-1) will have one intercept and one coefficient for the IV. $\endgroup$
    – Peter Flom
    Jan 11, 2013 at 18:17
  • 1
    $\begingroup$ @PeterFlom Oh yeah I remember that paper. You posted it on Quora. I think it is time to come up with a paper with R code. I might do that. $\endgroup$
    – Koba
    Apr 16, 2014 at 16:36
  • $\begingroup$ That would be cool $\endgroup$
    – Peter Flom
    Apr 16, 2014 at 21:04

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.