topic Re: Chi Square in SAS Procedures

Chi Square

yael — Tue, 09 Aug 2016 11:44:23 GMT

Hello

I intend to use PROC MEANS in order to get details (average, median, std, max, min) on 3 seperate groups. Is there any possibility

to check the difference between the groups using chi-square in this procedure? because I did not find it

Thanks a lot!

Re: Chi Square

Reeza — Tue, 09 Aug 2016 13:11:32 GMT

Chi square is via proc freq

Re: Chi Square

Rick_SAS — Tue, 09 Aug 2016 15:17:23 GMT

The chi-square test in PROC FREQ tests whether the proportions (or counts) in each cell of a one-way analysis of a discrete variable are significantly different. Is that your situation?

For a continuouis response variable, analysts often use PROC GLM to compare mean values across three or more groups. The GLM output includes an OverallAnova table that shows the result of an F test which indicates whether the means of the groups are significantly different.

Re: Chi Square

yael — Wed, 10 Aug 2016 05:55:26 GMT

Hello @Rick_SAS

Thanks for your reply and also for the explanation

I actually need the chi-square test to test whether the proportions in a discrete variable are significantly different. But the information about the PROC GLM is important to me for my next step.

Thanks a lot!

Re: Chi Square

yael — Thu, 11 Aug 2016 05:59:05 GMT

Hello @Rick_SAS

Following our correspondence above I would like to ask for your help,

I am trying to see if there is any difference between 3 groups of continouse variable. So I used PROC FREQ with Chi Square and got

a long list divided by 3 groups (it probably took every value in this list and indicated the frequency values). But more important is that I got the following message:

WARNING: 100% of the cells have expected counts less

than 5. Chi-Square may not be a valid test.

Effective Sample Size = 3986

Frequency Missing = 828

WARNING: 17% of the data are missing.

I checked and there is NO MISSING DATA but there is missing data if we divide it to 3 groups (attached an example)

First, I hope my description with my broken English is clear, BUT how can I solve it?

Thanks a lot!!!

Re: Chi Square

Rick_SAS — Thu, 11 Aug 2016 09:27:18 GMT

As I said in my previous message, PROC FREQ is for discrete response, typically counts. If you have a continuous response, you need to use a different procedure such as GLM.

Re: Chi Square

yael — Thu, 11 Aug 2016 10:52:40 GMT

Sorry! You are right, I forgot it