It is commonly used as a post-hoc test for the Kruskal-Wallis test, but can also be used as a post-hoc test for a Friedman test, or Cochran's q test. There might be more I'm not aware of. Usually the question is the other way around (i.e. I'm doing this non-parametric test, which post-hoc should I use). Hope this still helps.
(un)fortunately there is no law on how to report test results, and journals, schools, etc. will probably all have their preferences. An often heard APA-style is actually tricky, since the APA publication manual only shows examples for a few popular tests. In general they describe to give the reader 'sufficient' information, but that is of course a subjective term. I haven't checked their latest edition (7th) but doubt it will mention the Dunn test. In the end the Dunn test uses a normal distribution and a Z-value, so perhaps report it like you would a z-test and mention Dunn-Bonferroni. An example of the Z-test in APA style can be found easily with Google, but here is one result: courses.washington.edu/psy315/pdf/APA_format.pdf An alternative would be to search in Google scholar for an article that used a dunn-test and you have access to, to see how they report it. Hope this helps.
It can be that the KW test is significant, non of the comparisons in the post-hoc test are. This could be in a post-hoc test due to a correction being applied or even something known as Simpsons paradox. All pairwise comparisons resulting in 0.999 does seem a bit unusual. Perhaps forgot a '1 - ' somewhere?
Oops, will have to fix that in the future. Luckily though if you look at 5:16 you should see the full formula in the formula bar (on top) =SUMIF(A:A; J12) for the Rank sum. The ni formula is shown at 5:36. Hope this made it clear.
@@stikpet The equation =SUMIF(A:A;J12) seems to be the formula for ni. You talk about correcting the formula for Rank sum but we cannot see the new formula due to the video freezing. Based on the example file the formula should be =SUMIF(A:A; J12; C:C)
The Dunn test is used as a post-hoc analysis for comparing pairwise different categories. This is only relevant for situations where there are more than 2 categories. With the Mann Whitney U test there are only two categories, so there is no need for the Dunn test, and with the Wilcoxon signed rank test there are only two variables as well, so again no need for the Dunn test. For the Friedman test it is possible to use a similar approach. SPSS for example does this stats.stackexchange.com/questions/175441/friedmans-test-is-very-significant-but-its-post-hoc-comparisons-spss-are-not and I'm working on a video myself on how to do this in Excel, but haven't got the time to do so yet.
stikpet thanks for your reply..what is suitable post hoc for wilcoxon and mann whitney ? Or using the mean and standard deviation and z score is enough ?
There is no need for any post-hoc test after a Wilcoxon or a Mann-Whitney U test. You might want to check peterstatistics.com/CrashCourse/4-TwoVarPair/OrdOrd/OrdOrdPair3.html on the Wilcoxon test. The Mann-Whitney U test is only used if you have two categories. A post-hoc test is only used if you have more than 2 categories.
Thank you for this video! May I ask what are the non parametric tests which dunn test can be performed?
It is commonly used as a post-hoc test for the Kruskal-Wallis test, but can also be used as a post-hoc test for a Friedman test, or Cochran's q test. There might be more I'm not aware of. Usually the question is the other way around (i.e. I'm doing this non-parametric test, which post-hoc should I use). Hope this still helps.
@@stikpet you're such a great help! Thank you so much!
no problem, you're welcome. Good luck with your research/analyses/study.
I have a Z-value of 1.27 and Significance of 0.21 - what can I conclude from this? Thank you!
Have a look at peterstatistics.com/CrashCourse/3-TwoVarUnpair/NomOrd/NomOrd3b.html to see how to interpret the results. Hope this helps.
Hi, thank you for your video. Is there a formal way of writing the Dunn's test result?
(un)fortunately there is no law on how to report test results, and journals, schools, etc. will probably all have their preferences. An often heard APA-style is actually tricky, since the APA publication manual only shows examples for a few popular tests. In general they describe to give the reader 'sufficient' information, but that is of course a subjective term. I haven't checked their latest edition (7th) but doubt it will mention the Dunn test.
In the end the Dunn test uses a normal distribution and a Z-value, so perhaps report it like you would a z-test and mention Dunn-Bonferroni. An example of the Z-test in APA style can be found easily with Google, but here is one result: courses.washington.edu/psy315/pdf/APA_format.pdf
An alternative would be to search in Google scholar for an article that used a dunn-test and you have access to, to see how they report it.
Hope this helps.
@@stikpet Oh Fantastic. You are a star! You just saved my thesis!
@@kasanical glad to help. Good luck finishing your thesis.
@@stikpet Thank you
Hi this was very usefull! Do you know what it means when my kruskal-wallis gives p
It can be that the KW test is significant, non of the comparisons in the post-hoc test are. This could be in a post-hoc test due to a correction being applied or even something known as Simpsons paradox. All pairwise comparisons resulting in 0.999 does seem a bit unusual. Perhaps forgot a '1 - ' somewhere?
hi your video freezes at ~ 5:30 when you are inputting the formula for the rank sum. Could you please post it in a comment? Many thanks
Oops, will have to fix that in the future. Luckily though if you look at 5:16 you should see the full formula in the formula bar (on top) =SUMIF(A:A; J12) for the Rank sum. The ni formula is shown at 5:36. Hope this made it clear.
@@stikpet The equation =SUMIF(A:A;J12) seems to be the formula for ni. You talk about correcting the formula for Rank sum but we cannot see the new formula due to the video freezing. Based on the example file the formula should be =SUMIF(A:A; J12; C:C)
@@emilyahn1893 Thanks Emily. That should indeed be it.
Trying to look at the macro file for the UDF, it's telling me the macro file is locked, was this intentional?
yes, it's password protected. The macro simply follows the same steps as described in the video.
Hi Can dunn' s test be applied in wilcoxon sign rank and u mann witney and friedman test ?
The Dunn test is used as a post-hoc analysis for comparing pairwise different categories. This is only relevant for situations where there are more than 2 categories. With the Mann Whitney U test there are only two categories, so there is no need for the Dunn test, and with the Wilcoxon signed rank test there are only two variables as well, so again no need for the Dunn test.
For the Friedman test it is possible to use a similar approach. SPSS for example does this stats.stackexchange.com/questions/175441/friedmans-test-is-very-significant-but-its-post-hoc-comparisons-spss-are-not and I'm working on a video myself on how to do this in Excel, but haven't got the time to do so yet.
stikpet thanks for your reply..what is suitable post hoc for wilcoxon and mann whitney ? Or using the mean and standard deviation and z score is enough ?
There is no need for any post-hoc test after a Wilcoxon or a Mann-Whitney U test. You might want to check peterstatistics.com/CrashCourse/4-TwoVarPair/OrdOrd/OrdOrdPair3.html on the Wilcoxon test. The Mann-Whitney U test is only used if you have two categories. A post-hoc test is only used if you have more than 2 categories.
Your mouse clicking is SO annoying and distracting! It is about 10X louder than your voice haha!
yeah I know. Didn't realise it until much later. Sorry. Hope the video still helped. You can probably still follow along even on mute :D.