1
$\begingroup$

My dependent variable has 4 categories, but when I run the multinomial logistic regression using the package nnet with function multinom the results only show 3 categories.

I've tried changing the category numbers from 0,2,3,4 to 1,2,3,4, and also tried using names instead of numbers for the categories but it still wont show all 4 categories in the results.

Also, when I changed the categories to names instead of numbers, the resulting p values for each category drastically changed. Why is this? The p values were acquired using these commands

z <- summary(siglm)$coefficients/summary(siglm)$standard.errors
p <- (1 - pnorm(abs(z), 0, 1)) * 2
p
$\endgroup$
2
  • 3
    $\begingroup$ The missing category is the reference category. All the coefficients are interpreted with reference to that category. $\endgroup$
    – RJ-
    Commented Jan 20, 2016 at 6:52
  • $\begingroup$ Ah, OK, thank you! How do I get the p-value for that reference category? $\endgroup$
    – Nieve K
    Commented Jan 20, 2016 at 8:15

1 Answer 1

2
$\begingroup$

The missing category is the reference category. All the coefficients are interpreted with reference to that category.

You can change the reference category and run the same model to get statistics related to the reference category. By default the first factor level is the reference category.

$\endgroup$
6
  • $\begingroup$ Thank you for your answer, how do I change the reference category? Do I just change its number? Also if the reference category is changed, wouldnt that result in the p-values of the others being different, since they are related to a different category? $\endgroup$
    – Nieve K
    Commented Jan 20, 2016 at 8:57
  • 1
    $\begingroup$ in R, you can look it up for the proper syntax by doing factor $\endgroup$
    – RJ-
    Commented Jan 20, 2016 at 19:54
  • $\begingroup$ Thanks, is there a website that explains how to do this and how this works? Im a R beginner and this is very confusing for me :) $\endgroup$
    – Nieve K
    Commented Jan 20, 2016 at 21:58
  • 1
    $\begingroup$ ats.ucla.edu/stat/r/dae/mlogit.htm $\endgroup$
    – RJ-
    Commented Jan 20, 2016 at 21:59
  • $\begingroup$ cookbook-r.com/Manipulating_data/… $\endgroup$
    – RJ-
    Commented Jan 20, 2016 at 22:00

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.