turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Data Mining
- /
- Score or points allocations after building logisti...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-10-2009 09:55 PM

I came up with a logistic regression using EG. I transformed the probability to score using the following equation: score=Log(odds) * factor + offset

Now, I need to come up wtih the points or score allocation. For an example, let's say if age < 30, then the account will get points of 7. For age >=30, it will get 12 points. I heard I have to use weight of evidence. I'm wondering if there is a SAS option to do that quickly.

Now, I need to come up wtih the points or score allocation. For an example, let's say if age < 30, then the account will get points of 7. For age >=30, it will get 12 points. I heard I have to use weight of evidence. I'm wondering if there is a SAS option to do that quickly.

Accepted Solutions

Solution

07-07-2017
03:11 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-21-2010 01:16 PM - last edited on 07-07-2017 03:12 PM by DougWielenga

All Replies

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

06-15-2010 01:25 PM

My colleagues and I built a spreadsheet to do that for us. Basically you build a formula that calculates the score points off of your estimate. For example,

-XX*LOG(EXP(estimate),2) where the number XX is the value for which your score doubles in odds.

-XX*LOG(EXP(estimate),2) where the number XX is the value for which your score doubles in odds.

Solution

07-07-2017
03:11 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-21-2010 01:16 PM - last edited on 07-07-2017 03:12 PM by DougWielenga

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-21-2010 01:24 PM

What would be great is if EM could just take the output you get from a regression node and create a scorecard without having to go through the regression steps again.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-21-2010 01:29 PM

Yep. See Wayne's answer about the credit score package. Also look for the next update to the software. Thanks.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

08-12-2012 07:59 AM

1. The exercise per se is not a modeling process. Just a transformation, if you would. Or a mapping exercise.

2. Score-carding generally is industry-specific. There is general-purpose scorecard and credit application specific score card. As Wayne indicated, for credit scoring business, it may be fairly straightforward to either use EM credit score application or run the log transformation as indicated in the spreadsheet (through Score node or just data steps). EM, however, has a much broader user base than credit risk modeling. Many users do not use log transformation to map probability scores to card allocation. If you google, you should find people focusing on all kinds of scaling exercises involving score card

3. If you really need to see a custom node (icon) in EM to conduct a focused, custom task for yourself (I know some users like to click button to get things done), consider extending EM node, although I personally believe EM SAS Code node should be easy kill for this exercise.

4. If you goal is for model performance comparison, you can use Model Import node in EM to get the probability score into EM and combine it into Model Comparison node in later steps.