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
- /
- beta estimate of reference group

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

11-18-2014 03:09 PM

Hi all,

One question,

when i run a logistic regression with a binned variable (age in 3 categories) and i use the 1st group as reference group, which beta estimate does it get?

because i get an estimate for the other two groups but not for that one or am I mistaken?

The reason i am asking cause i would like to have the according estimates so as to calculate the probability myself but i will be missing out on cases where the id falls into the 1st bin.

Regards,

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

11-18-2014 03:27 PM

You don't get an estimate for that group. Make sure to specify PARAM=REF in your model.

IE. If you have three variables, that gets recoded to 2 binary variables. When both are 0 then the third/reference level is true.

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

11-18-2014 03:36 PM

Beta estimate for reference group will not be used for probability calculations. Because there will be no dummy variable for this one in the model.

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

11-18-2014 03:47 PM

ok so how can an id which falls into that bin get a probability then?

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

11-18-2014 03:56 PM

If you have no other variables it's the intercept, if you have other variables its the equation without that variable.

Lets assume you have a variable AGE with 3 categories: 1, 2, 3. Reference level is set to 1.

X1=1 if age=2

X1=0 if age=1,3

X2=1 if age=3

X2=0 if age=1,2

Model = Intercept + B1*X1 + B2*X2

If Age=1 then

Model = Intercept

If Age=2 then

Model = Intercept + B1*X1

If Age=3 then

Model = Intercept + B2*X2

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

11-18-2014 04:15 PM

You calculate it from the intercept(parameter).

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

11-18-2014 03:56 PM

meaning an id with ref 1 for age and 0 for gender what predicted probability will it get?

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

11-18-2014 04:11 PM

aha ok, sure that makes sense, thank you very much for the reply and example it was really helpful, by the way you wouldnt happen to have any good example on creating the KS chart in excel?

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

11-18-2014 04:24 PM

What is a KS chart?

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

11-19-2014 04:29 AM

Kolmogorov Smirnof chart