5
$\begingroup$

Am I understanding this right? I have run a negative binomial regression on overdispersed count data (Y is number of litter items found, and X is the distance to the shoreline), in SPSS. The omnibus test result is highly significant but I'm quite puzzled when interpreting the coefficient in the parameter estimate: enter image description here

If I understand correctly, for every increase in X (i.e. the independent variable), I will have a decrease in log count in the Y (i.e dependent variable). The result of this is significant so I can trust it. When I look at the coefficient B (-,042) I understand it being in natural logarithm form, so I could interpret it as a decrease of 3,17 quantity of litter? (that is changing -,042 in natural log form to a standard numeric form).

Improve this question
$\endgroup$

1 Answer 1

Reset to default
7
$\begingroup$

I'm not sure how you are getting the 3.17. Taking the exponent of -0.042 returns 0.959. Thus, for every one-unit increase in distance, the number of litters decreases by 4% [(1 - 0.959)x100=4].

Improve this answer
$\endgroup$
3
  • 1
    $\begingroup$ thanks for your input. Well, I must have made an error somewhere then, my bad! $\endgroup$
    – Tim56
    Commented Jul 22, 2020 at 12:03
  • 1
    $\begingroup$ so here just which exponent did you consider? $\endgroup$
    – Tim56
    Commented Jul 22, 2020 at 14:58
  • 2
    $\begingroup$ The log link used in negative-binomial regression is the natural log, which has base = e. Thus, the antilog is e^x. In most software programs (and calculators), this is exp(). $\endgroup$
    – dbwilson
    Commented Jul 22, 2020 at 16:12

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.