Yes I have finally put the testing to rest. I may give the Friedman test a go later on, but on SPSS. A colleague of mine told me are attemtping to conduct the Friedman test on SPSS.
But thank you for all the help and good advices.
I am also getting the error, "Blocking ignored because cell counts were unequal" when I try to do a simple ANOVA, despite the fact that all my blocks have an equal number of observation. Were you ever able to figure this out? Thanks!
I am afraid not. I gave the test up some time ago. But I may look into it again as I have some new data, soon, that will be analyzed statistically. I will update on this issue if Friedman is of interest again.
How can I run that Friedman's test for a repeated-measures design, please?
Let me know the step procedures.
I asked the same question and JMP support replied this, which works
JMP has many nonparametric tests built in. The Friedman’s test is not directly available in JMP, but you can transform your data into ranks in JMP to create this test as just a 2-way ANOVA on the ranks (using the blocking factor and your other factor as the two additive factors). There is a blog about this (and arguing, at the end, why you might not want Friedman’s test in some cases) here:https://community.jmp.com/t5/Discussions/Friedman-test-on-JMP/td-p/5974.
It is a problem if you have ties, which may be why your data are not normal in the first place. People often forget that alhtough non parametric tests don't assume that the data are normal, these tests still have assumptions. Hope this helped
Thank you for your reply and link. I took a look at those posts, but as I understood that problem didn't refer to a repeated-measures design, did it?
When we face a repeated-measures design, in general, the stats procedures change due to the peculiarity of that design.
Jeff (@Jeff_Perkinson), do you have any suggestion to solve this problem, please?
Thank you very much.
Once you have the ranks, you are essentially doing a repeated measures on the ranks - so you are transforming the data and conducting the RM analysis on the ranks. That is basically what a Friedmans test is doing under the hood
You mean using Fit Model for ranks in order to perform the RM analysis. However, in the last posts I saw using Fit Y by X for calculating the blocks. Perhaps, I need to do it first and then go to the Fit Model. As I understand, RM can be evaluated in the Fit Model only.
I realise this is an old discussion, but I found it most helpful. I have been able to successfully implement it but have the following question:
For evaluating the impact of both variables on the outcome variable, I assume one must test each separately, using ranks created using both but alternating them between blocking and dependent roles. However, how does one determine interaction between these two variables? Luckily I was able to transform my data successfully and perform a two way ANOVA using the Fit Model platform. Is this a limitation of the Friedman test?