Help with ordinal logistic regression

[modeller2017]

Hi ,

Can anyone pelase help me with following question I am trying to solve:

Description of my problem:

I have the following "target variable": 1: Growth 2. Stable 3. Decay. 

The distribution of each is:

1. Growth=4%

2.Stable=88%

3.Decay=8%

Since, this is situation of imbalance target rate I have taken 10% of the stable to model.

Now I have the model results and the confusion matrix.

Issue:

When I try to try back the complete sample, I see the values of predicted probabilities get disturbed.

Can anyone please suggest how to score the complete sample in ordinal regression when we model only a sample of it.

[Reeza]

How are you scoring your dataset? 

Have you considered a prior probability instead of reducing event rate? 

Otherwise, perhaps this is helpful

http://support.sas.com/kb/22/601.html

[modeller2017]

Hi Reeza,

Thank you for your reply. I have read this article but have trouble implementing for ordinal case. Can you please advise how should I implment for ordinal case?

Thanks a lot again!!