I would like to use a Friedman test to compare between them more than two dependent samples. In your opinion, is a sample of 30 animals enough to use this non parametric test?
Many thanks for any hint.
I'm not a specialist in the field of the question posed here but think that the paper at the following link might be useful in designing animal experiments:
http://www.3rs-reduction.co.uk/assets/applets/Festing_Altman.pdf
Depends on the margin of error or variation that you want to accept. The standard error is proportional to 1/root n where n is the number of subjects, experiments etc. So your expected standard error is around 18% with 30 subjects. To bring this down to 10% you'll need 100...
A couple of notes on Friedman test:
1) Friedman test can only be used for unreplicated complete block design.
2) If there are fewer than 5 groups, some people recommend Quade test as an analogous test with more power.
I'm not a specialist in the field of the question posed here but think that the paper at the following link might be useful in designing animal experiments:
http://www.3rs-reduction.co.uk/assets/applets/Festing_Altman.pdf
Hello Sofia. You may find interesting Thom Baguley's blog post entitled "Beware the Friedman Test!" Note especially the section headed with "Why the Friedman test is an impostor". HTH.
https://seriousstats.wordpress.com/2012/02/14/friedman/
- Badges
- Science topic
- Similar topics
- Medicine
- Life Sciences