Programming the statistical procedures from SAS

Statistical Test when cross table has less than 5 cases for majority of the cells

Accepted Solution Solved
Reply
Frequent Contributor
Posts: 92
Accepted Solution

Statistical Test when cross table has less than 5 cases for majority of the cells

Dear Team, 

 

i am working on a survey data and need to run a chi square test to see if there is any relation between gender and rating. 

 

As one of the assumption of Chi Square is : Sample Size: each particular scenario (cell) must have at least five cases. However as i have very less data points hence this assumption is violated. 

We can run a fisher exact test in this scenario though while researching i came across a literature from "Wai Wan Tsang and Kai Ho Cheng" from the Univeristy of The University of Hong Kong" that we can still run a chi square test. 

 

I am a bit confused and while searching this on researchgate. net this option was suggested hence checking on the forum if this is still a valid test to conduct. 

 

Regards, Shivi


Accepted Solutions
Solution
‎06-27-2016 10:05 AM
SAS Super FREQ
Posts: 3,307

Re: Statistical Test when cross table has less than 5 cases for majority of the cells

Yes, it sounds like exact tests would be appropriate. For an overview of exact tests in PROC FREQ, see "Exact tests in PROC FREQ: What, when, and how."

 

If you have many dozens of cells, a full exact test might be prohibitively expensive. In that case, you can use Monte Carlo approximations to the exact test by adding the MC option to the EXACT statement:

exact pchi / MC;     /* Monte Carlo approximation to exact chi-square test */

View solution in original post


All Replies
Respected Advisor
Posts: 4,606

Re: Statistical Test when cross table has less than 5 cases for majority of the cells

You can request exact statistics for the likelihood ratio chi-square test (if that what your reference is about) with EXACT LRCHI;

PG
Solution
‎06-27-2016 10:05 AM
SAS Super FREQ
Posts: 3,307

Re: Statistical Test when cross table has less than 5 cases for majority of the cells

Yes, it sounds like exact tests would be appropriate. For an overview of exact tests in PROC FREQ, see "Exact tests in PROC FREQ: What, when, and how."

 

If you have many dozens of cells, a full exact test might be prohibitively expensive. In that case, you can use Monte Carlo approximations to the exact test by adding the MC option to the EXACT statement:

exact pchi / MC;     /* Monte Carlo approximation to exact chi-square test */
Frequent Contributor
Posts: 92

Re: Statistical Test when cross table has less than 5 cases for majority of the cells

Thanks Rick for the solution. I think this works perfect. Also thanks for sharing the blog link.

☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 3 replies
  • 178 views
  • 3 likes
  • 3 in conversation