topic Re: Displaying residuals in Chi-Squared contingency tables? in Discussions

Displaying residuals in Chi-Squared contingency tables?

Poparad — Fri, 09 Jun 2023 00:42:28 GMT

When performing a Chi-Square test, is there a way to display the cell residuals in the resulting contingency table? I haven't been able to find any explanations online for how to do it, so is this simply something that JMP can't do?

Re: Displaying residuals in Chi-Squared contingency tables?

dale_lehman — Sat, 27 Nov 2021 14:00:28 GMT

I assume by "residuals" you mean the deviations of the cell values from what would be expected under the hypothesis of independence. If so, then it is available from the drop down menu at the red arrow for he Contingency Table. You can get the expected cell values, the deviation from these expected values, and the cell chi-squared value.

Re: Displaying residuals in Chi-Squared contingency tables?

Poparad — Sat, 27 Nov 2021 22:11:03 GMT

No, residuals are calculated as (observed - expected) / sqrt(expected). If the value is greater than 2 or less than -2, then that cell is a significant contributor to the variation. Basically, it's the post-hoc test for the Chi-square test. I can do it by hand in Excel since it's an easy calculation and JMP does display the observed and expected values in the table, but it's time consuming and tedious. I was hoping for a quick, automatic way to generate them.

Re: Displaying residuals in Chi-Squared contingency tables?

dale_lehman — Sat, 27 Nov 2021 22:20:02 GMT

The Cell Chi^2 is the next choice on the drop down menu after the "Deviation."

Re: Displaying residuals in Chi-Squared contingency tables?

Mark_Bailey — Mon, 29 Nov 2021 19:01:26 GMT

And the deviation is the square root of the cell chi square value. So you could also compare the cell chi square to 4 instead of comparing the deviation to 2.

I am not sure about the validity of this 'rule of thumb' (|deviation| > 2) but you see to have experience using it successfully.