- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hello friends,
I have a data set of survey results that were collected from students. I have a 2x3 table and want to see the association between two variables. I compiled the data and inputed it into SAS and attempted to use Two-Way Anova for comparison. All is going according to plan but when i ran the interaction from proc glm for two variables, SAS was unable to give me a p-value and it was left blank. If I to understand this as meaning the p-value is so tiny, it is nonexistant?
Pengy
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes, it sounds like you have count data, so a chi-square test is more appropriate based on what you're saying.
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
You say 2x3 table but then Anova which makes me wonder. What type of data did you collect, categorical or continuous?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Well my data was a question that the students responded to. I either marked them down as "Yes, No, or Undecided" and separated answers based on gender which produced the 2x3 table. I assumed Two-way Anova was the right course of action but maybe a chi-square test would be more fitting?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes, it sounds like you have count data, so a chi-square test is more appropriate based on what you're saying.
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
My only concern is additionally, I need to determine the expected cell counts and test to see if gender and survey response are associated/have interaction. Would a chi square test be able to show interaction? Maybe a box plot cold be beneficial for that?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Look at proc freq. A visual display of data is always good, but box plots are for continuous data.
Here's an example that sort of matches up what you're trying to do.
Base SAS(R) 9.2 Procedures Guide: Statistical Procedures, Third Edition
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thank you for assisting, Reeza
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
One last thing. There now is a way to do Mosaic plots in Proc Freq. See link to Rick Wicklin's blog