Even when your analysis is set up perfectly in Prism, the results are only valid if your data meet the fundamental assumptions of the chi‑square test.
The p-value answers the core question: If the null hypothesis were true (no association), what is the probability of seeing an effect this large or larger?
: Entering percentages, averages, rates, or scaled values will completely invalidate your results.
– If your contingency table has more than two rows or columns and the overall chi‑square is significant, use pairwise z‑tests (with a Bonferroni correction) to identify which specific categories differ from one another. Prism can perform these through the “Multiple comparisons” option in the contingency table analysis. chi square graphpad verified
This point cannot be emphasized strongly enough. Prism will accept entries such as “42%” and “58%” without giving you any warning – but the chi‑square results will be . You must enter the actual frequencies (the raw counts). For example, if 12 patients out of 30 in a group experienced an event, you enter “12” in the “Event” cell and “18” in the “No Event” cell – not “40%” anywhere in the table. Garbage in, garbage out is the rule here.
Once your data are entered, go to the top menu and click . In the dialog that appears, choose Contingency table analyses from the list, then select Chi‑square (and Fisher’s exact) test . Click OK .
): There is no association between the variables (for contingency tables) or the observed data follows the expected distribution (for goodness-of-fit). Alternative Hypothesis ( Hacap H sub a Even when your analysis is set up perfectly
Input your observed values into the rows (e.g., Treatment, Control) and columns (e.g., Improved, Not Improved).
A common rule of thumb is that all should be 5 or greater. In a 2×2 contingency table, this means that the number of subjects Prism calculates for each cell under the null hypothesis (these are displayed in the results) should be at least 5 in all four cells.
: Reject the null hypothesis. There is a statistically significant association between your variables. – If your contingency table has more than
Select (or choose Fisher’s exact test if your sample size is very small, typically when cell counts are below 5).
If 50% of 20 mice survived, do not enter 0.50 or 50 into Prism as a continuous value. Enter the absolute count: 10 .