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
- /
- Chi Square test or logistic regression for the kin...

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

12-01-2014 01:37 PM

I was wanting some assistance with knowing how to create a contingency tables or find chisquare statistics for an independent variable such as smoking status coded as 0,1 ( yes, no) and another variable that has more than two levels such as education level,

0=None, 1=Primary, 2=Secondary, 3=Technical, and 4= Four year college university and beyond.

I know that if both variables had two levels I could do a chi square test, but was not sure if there were two variables and one had more levels as to what to do? I would like to see the significance between level of education and smoking status.

Thank you

Accepted Solutions

Solution

12-01-2014
02:20 PM

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

Posted in reply to LALRJ15

12-01-2014 02:20 PM

You can still do a chi square test if one variable has more than two levels.

If the cell counts get low you may want to use a Fisher's Exact test instead.

All Replies

Solution

12-01-2014
02:20 PM

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

Posted in reply to LALRJ15

12-01-2014 02:20 PM

You can still do a chi square test if one variable has more than two levels.

If the cell counts get low you may want to use a Fisher's Exact test instead.

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

Posted in reply to Reeza

12-03-2014 04:46 AM

Thank you!

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

Posted in reply to LALRJ15

12-03-2014 05:49 AM

Also I had another question, if I wanted to see if there are significant differences in the levels of one categorical variable, what type of test would I conduct?

Thank you.

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

Posted in reply to LALRJ15

12-01-2014 04:21 PM

Since your predictor (education) is ordinal, you would generally want to do a trend test. This can be done without fitting a model using the TREND option in PROC FREQ. For this and other ways, see the **Trend test for ordered alternatives **item in the list Frequently Asked-for Statistics (FASTats).

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

Posted in reply to StatDave_sas

12-03-2014 04:51 AM

Thank you for your help. I want to understand this correctly. You conduct a chi square test to see if there is an association between your independent and dependent ( response ) variable first. If it is found to be significant, this means an association is supported between these two variables? To investigate deeper into as where the significance is observed, in what level it is observed, let's say level of education in the previous example, would I then have to do a logistic model to see the significance among each level in correspondence to the dependent variable?