I have been working on ordinal regression in SPSS using two different group of variables but the same Y. After I removed the non-significant variables in each equation I have ended up with two equations explaining the same Y. What I am interested now is to find which equation explains Y better.
I can understand that I cannot compare Deviances because I am using different variables. What I thought was to use the coefficients of the models and create another column for each model with the predicted Y. Then I can compare the correlation between Y->Y of model 1 and Y->Y of model 2. However, can I create a single predicted value for Y or just the probability for each of the 5 possible responses?
Lastly, I am not sure how i can interpret and use the output of SPSS in ordinal regression to calculate the predicted Y for each row. Are we using the exponential of the coefficient? I am kind of confused on how to proceed.
my Y has 5 ordered categories.
Output model 1:
Threshold [Y=1] 2.919
Location X1 .316
Link function: Logit