I have a model developed in R using the polr() function in the MASS package. The model is an ordinal regression with a complementary log-log link (method=cloglog). The model uses four predictors called X1, X2, X3 and X4 and predicts membership of four ordinal outcome classes called A, B, C and D. The coefficients and intercepts are given below.

Coefficients: X1=1.04; X2=-0.59; x3=1.85 ; X4=-.02

Intercepts: A|B=-5.19; B|C=-4.03; C|D=-2.92

I want to calculate the model's predicted probabilities in mapping software (outside R) with X values extracted from raster layers. To do this though I need to know the actual formula to apply to the raster layers.

The help for the polr() function says "The complementary log-log link is the increasing function F-1(p) = log(-log(1-p))" (see attached image for clarity, I had trouble with the formatting).

polr() cloglog link function

Unfortunately this level of maths is above my head. I am hoping that someone can give me the text form of the above formula with my coefficients and X values included (analogous to P(A) = 5.19 + 1.04*X1 - 0.59*X2 +1.85*X3 -.02*X4). I need the predictions in response (probability) scale and not link scale. Many thanks for any help.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.