# Predicted Probabilities for Ordinal Regression

#### beginner1.mat

##### New Member
Hi,

I ran an ordinal regression with 7 levels in SPSS. I am using age, sex and weight as input/predictor variables. I would like to understand what SPSS is doing to get the estimated probabilities that are exported to the data columns after running the analysis (i.e., EST1, EST2, EST3,EST4, EST5, EST6, EST7). When you sum these up, they equal 1 and the predicted value is the one with the largest percentage. I tried using data in the "Parameter Estimates" table to calculate the probabilities, but am not getting the same results. This is what I used so far to try to calculate the estimated cumulative probabilities:

EST1=1/(1+EXP(-1*(Beta1-Sex_Coef*Sex - Age_Coef*Age - Weight_Coef*Weight)));
EST2=1/(1+EXP(-1*(Beta2-Sex_Coef*Sex - Age_Coef*Age - Weight_Coef*Weight)))-EST1;
etc...

Beta1= "Estimate" from the Threshold rows
Coef = "Estimate" from the Location rows, for the specific input/predictor variables.

Can anyone confirm if my equation is correct? Any help would be appreciated!

Thank you!