Use this when you want to compare two categorical variables to see if they are related (e.g., "Is treatment outcome (cured/not cured) associated with the type of drug administered?").
GraphPad Prism supports three primary versions of the Chi-square analysis depending on your research design:
Once your table is filled with counts:
: Also called the Cochran-Armitage test, this is used when you have two columns and multiple rows arranged in a natural, ordered sequence (like increasing doses of a drug). 2. Step-by-Step Guide to Running the Test Follow these steps to perform the analysis in the software:
No expected cell count should be zero, and fewer than 20% of cells should have expected counts <5. If these fail, consider Fisher’s exact test (Prism automatically offers this for 2x2 tables).
If you have a column of "Group" and another column of "Outcome":
For $2 \times 2$ tables, some statisticians prefer Yates' continuity correction. Prism offers this as an option in the parameters window (" Yates' correction (recommended)"). In modern statistics, Fisher’s Exact Test is generally preferred over Yates' correction for small samples, so sticking with the default or Fisher's is usually best.
A number-heavy results table is less impactful than a graph.
The parameters dialog is simple. You will rarely need to change the default settings, but note these options:
How to do a Chi square or Fisher's exact test in GraphPad Prism
Use this when you want to compare two categorical variables to see if they are related (e.g., "Is treatment outcome (cured/not cured) associated with the type of drug administered?").
GraphPad Prism supports three primary versions of the Chi-square analysis depending on your research design:
Once your table is filled with counts:
: Also called the Cochran-Armitage test, this is used when you have two columns and multiple rows arranged in a natural, ordered sequence (like increasing doses of a drug). 2. Step-by-Step Guide to Running the Test Follow these steps to perform the analysis in the software:
No expected cell count should be zero, and fewer than 20% of cells should have expected counts <5. If these fail, consider Fisher’s exact test (Prism automatically offers this for 2x2 tables).
If you have a column of "Group" and another column of "Outcome":
For $2 \times 2$ tables, some statisticians prefer Yates' continuity correction. Prism offers this as an option in the parameters window (" Yates' correction (recommended)"). In modern statistics, Fisher’s Exact Test is generally preferred over Yates' correction for small samples, so sticking with the default or Fisher's is usually best.
A number-heavy results table is less impactful than a graph.
The parameters dialog is simple. You will rarely need to change the default settings, but note these options:
How to do a Chi square or Fisher's exact test in GraphPad Prism