topic Re: Categorical Correlation in Discussions

Categorical Correlation

AT — Thu, 02 Aug 2018 20:33:36 GMT

Hi,

I like to do two categorical variables correlations. I know I have to do the Chi-Square analysis. The questions I have is how many data points I need to do this? Are there any minimum requirements?

Thanks

Re: Categorical Correlation

txnelson — Thu, 02 Aug 2018 21:23:19 GMT

The p Value you receive from a Chi Square, will be adjusted for the number of observations. Therefore, it will self adjust for your analysis. However, where Chi Square does fall down, is when you get too many cells where the "Expected" cell frequencies are below 5 that can make the calculations suspect. There is also some discussions on having at least 20 responses.

Re: Categorical Correlation

AT — Fri, 03 Aug 2018 00:46:46 GMT

Thanks Jim. I have a two level categorical variables and I guess you are sugggesting to have at least 20 data points for each level.

Re: Categorical Correlation

txnelson — Fri, 03 Aug 2018 01:12:10 GMT

That would be fine

Re: Categorical Correlation

Mark_Bailey — Fri, 03 Aug 2018 14:44:12 GMT

The general rule for the hypothesis test of association with Chi-square is nor more than 20% of the cells have an expected count under 5. JMP will warn you if this condition is discovered.

Re: Categorical Correlation

AT — Fri, 03 Aug 2018 16:19:09 GMT

Thanks. That's very helpful.