I have three sets of data, to which I want to apply Dunn's test. However, the test shows different results when performed in GraphPad Prism and R. I've been reading a little bit about the test here, but I couldn't understand why there is a difference in the p-values. I even tested in R all the methods to adjust the p-value, but none of them matched the GrapPad Prism result.

Below I present screenshots of the step-by-step in GraphPad Prism and the code I used in R.

enter image description here enter image description here enter image description here enter image description here


Day <- rep(1:10, 3)
FLs <- c(rep("FL1", 10), rep("FL2", 10), rep("FL3", 10))
Value <- c(0.2, 0.4, 0.3, 0.2, 0.3, 0.4, 0.2, 0.25, 0.32, 0.21,
           0.9, 0.6, 0.7, 0.78, 0.74, 0.81, 0.76, 0.77, 0.79, 0.79,
           0.6, 0.58, 0.54, 0.52, 0.39, 0.6, 0.52, 0.67, 0.65, 0.56)

DF <- data.frame(FLs, Day, Value)

Dunn <- DF %>%
  dunn_test(Value ~ FLs,
            p.adjust.method = "bonferroni",
            detailed = TRUE) %>%

enter image description here


1 Answer 1


I don't think the two programs are answering the same question.

In Prism, you specified that each row (data for each day) is a matched set, so a repeated measures analysis is performed. Because you specified a nonparametric test, it does the Friedman test with Dunn followup comparisons.

I am not very familiar with those R commands, but it doesn't look like you specified pairing or repeated measures. I think your R analysis is doing the Kruskal-Wallis nonparametric test (without pairing) with Dunn followup comparisons.

"Dunn's" test just means it corrects for multiple comparisons using what is often called the Bonferroni method (but it is more appropriate to attribute to Dunn). Dunn's adjustment can be done for many kinds of analyses, including repeated measures (paired) or not.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.