0
$\begingroup$

I am using SPSS. On my dataset I get a highly significant Friedman test(p<0.001) but post hoc pairwise tests with Dunn-Bonferroni (in SPSS this test is set as the default test after a significant Friedman test) are not significant (both adjusted an unadjusted p-values are not significant). I am evaluating how four different image reconstruction algorithms (4 different groups) are effecting image quality grading. I am mostly interested in the differences between each pair of groups. My question is if I should report the p-value from the Friedman test at all in my manuscript, or if I should only report the p-values from the post-hoc pairwise comparisons, or should I report both Friedman and post-hoc pairwise comparisons p-values? I am aware of the fact that the abovementioned test results are not surprising because the two tests answer different questions (Friedman compares entire/nested models and not groups), and that is not my question. My question is about your experiences of what you would report in the manuscript in this situation. Will the reviewers call my analysis wrong if I do not report the Friedman test? Will they object if I only provide the posthoc tests (since I am mostly interested in the comparisons between groups)? Looking forward to your reply. /Michael

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.