Friedman & Paired Comparisons (Durbin-Conover)

Discuss statistics related things
Post Reply
pstjacques
Posts: 2
Joined: Tue Jun 23, 2020 7:43 pm

Friedman & Paired Comparisons (Durbin-Conover)

Post by pstjacques »

I am conducting Friedman non-parametric test, followed by pairwise comparisons (Durbin-Conover). I was wondering if the p-values reported in the table are already adjusted for multiple comparisons?
User avatar
MAgojam
Posts: 421
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

Re: Friedman & Paired Comparisons (Durbin-Conover)

Post by MAgojam »

Hi, @pstjacques.

The p values shown in the table are obtained as:
P value adjustment method: none
Take a look at the attached screenshot, where I use the Rj module and this little r script.

Code: Select all

library(PMCMR)
options(digits=5)
y <- data.matrix(data[1:6])
friedman.test(y)

# P value adjustment method: none
posthoc.friedman.conover.test(y, p.adj="none")

# P value adjustment method: bonferroni
posthoc.friedman.conover.test(y, p.adj="bonferroni")
ScreenShot.png
ScreenShot.png (105.26 KiB) Viewed 13359 times
Cheers,
Maurizio
pstjacques
Posts: 2
Joined: Tue Jun 23, 2020 7:43 pm

Re: Friedman & Paired Comparisons (Durbin-Conover)

Post by pstjacques »

Thanks very much for your response Maurizio! I'm a newbie to R, so it's helpful to see how you have used the Rj Module here.
jpmaroco
Posts: 13
Joined: Fri Mar 24, 2017 4:24 pm

Re: Friedman & Paired Comparisons (Durbin-Conover)

Post by jpmaroco »

Dear Maurizio,
Can I jump into this?
Looking at the R script (in your print screen) one can see that the post-hoc test is `posthoc.friedman.conover.test(y=y, p.adjust="none")` from the PMCMR[/code] library. However, the test is named Durbin-Conover in jamovi. In he PMCMR there's also a Durbin test (for incomplete block designs that is the same as Friedman where the design is complete) and the Durbin post-hoc statistic in `posthoc.durbin.test` looks quite different from the Conover one.... So I am wondering why jamovi calls it Durbin-Conover (maybe it is because that both tests are the same when the design is complete?).
Warm regards,
João
User avatar
MAgojam
Posts: 421
Joined: Thu Jun 08, 2017 2:33 pm
Location: Parma (Italy)

Re: Friedman & Paired Comparisons (Durbin-Conover)

Post by MAgojam »

jpmaroco wrote:So I am wondering why jamovi calls it Durbin-Conover (maybe it is because that both tests are the same when the design is complete?)João
Hi, João.
Sorry I'm late, but I lost you (I'm old).
We can resent in SLACK, because we are a bit distant in time from the post.
My intervention here was what was going on under Jamovi's hood.
In SLACK there are the 'J guys' who can motivate the choice.

Uma saudação calorosa.
Maurizio
Post Reply