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
- /
- Stat Procs
- /
- Check categorical variable distribution

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
- Permalink
- Email to a Friend
- Report Inappropriate Content

08-04-2015 05:04 PM

I want to check categorical variable distribution over a period of time. For example, the distribution of variable district code has changed over a period of time ( Time 0 - Period of model Development, time1 - after one year).

I perform the following analysis for numeric variables.

Step 1 - Rank variable into 10 group (decile)

Step 2- (% of records based on variable in Scoring Sample (A) - % of records based on variable in Training Sample (B)) * In(A/ B)

Step 3 - Then sum up the scores in step2 on 10 groups

I'm little bit skeptical about the same analysis for categorical variables. I guess chi square analysis isn't correct technique to check it. Am i correct? What's the correct way?

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

Posted in reply to Ujjawal

08-04-2015 05:31 PM

Take a look at the the options in proc freq, there's a test for trends over time.

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

Posted in reply to Reeza

08-05-2015 12:27 AM

Which test?

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

Posted in reply to Ujjawal

08-05-2015 09:02 AM

This categorical variable is nominal or ordered ?

and how many levels does this variable have ?

If two level , you could try TRAND analysis.

tables var * time / trend cl ;

Otherwise , try MEASURE analysis.

tables var* Dose / measures cl;

Xia Keshan

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

Posted in reply to Ujjawal

08-05-2015 10:30 AM

TREND test

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

Posted in reply to Reeza

08-05-2015 10:46 AM

Trend test is only for 2*N matrix , if I was right . while MEASURE(Association test) is for N*M matrix.

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

Posted in reply to Ksharp

08-05-2015 10:50 AM

OP mentioned 2*N - Time before and Time after measurements.