0
$\begingroup$

I get an output of ordinal regression model using r. it has 3 levels but I can't understand how to interpret can you explain to me what is the meaning of 0/1 and 0/2?

enter image description here

here logic is the predicted variable....

                               Value Std. Error     t value      p value
sex                           -0.1713340  0.6552480  -0.2614796 7.937227e-01
nationality                    1.9440132  1.5624353   1.2442200 2.134186e-01
race                         -17.6845396  1.3415344 -13.1823233 1.109215e-39
religion                       6.4993120  0.7103028   9.1500582 5.690286e-20
place.of.origin               -0.1388703  0.5199239  -0.2670974 7.893942e-01
recidence.at.university.time  -0.3090342  1.2186897  -0.2535791 7.998208e-01
0|1                           -9.2347103  2.2702138  -4.0677713 4.746494e-05
1|2                           -7.8270152  2.3122740  -3.3849861 7.118186e-04

$\endgroup$
  • $\begingroup$ You need to investigate why race has a coefficient which is tending to $-\infty$. Check this site for separation. $\endgroup$ – mdewey Feb 16 at 16:43
0
$\begingroup$

Not sure what library you're using but I would assume $0|1$ to be the value of the constant term $\beta_{0[i=1]}$ for equation:

$$log~{Odds_{1|0}} = \beta_{0[i=1]} +\beta_{1}x_1 +\beta_{2}x_2 + ... +\beta_{n}x_n$$

and $1|2$ to be the value of the constant term $\beta_{0[i=2]}$ for equation:

$$log~{Odds_{2|1}} = \beta_{0[i=2]} +\beta_{1}x_1 +\beta_{2}x_2 + ... +\beta_{n}x_n$$

Due to the proportional odds assumption, $\beta_1, \beta_2, ..., \beta_n$ should be equal across all equations, that's why you only get one estimate for each of them. The constant term $\beta_{0[i]}$, however, will be different for ever base level $i$ of the ordinal response variable, and therefore has a unique estimate in every equation.

$\endgroup$
  • $\begingroup$ thank you very much but here I take the prediction variable has 3 levels 0, and 2 $\endgroup$ – Lahiru Sandaruwan Feb 16 at 16:01
  • $\begingroup$ For 3 levels your model will have 2 equations, the third one being the reference level. $\endgroup$ – Digio Feb 16 at 17:00

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.