Friedman Test vs. Durbin Test

MaximusLuminis

Hello,

in my analysis I am conducting a Repeated Measures ANOVA. Since the data is not normally distributed, I am using the Friedman Test. However, JASP defaults to the Durbin Test and does not give me the Friedman Test. The Durbin Test is typically used in incomplete block designs, but my design is not incomplete:

Participants came into the lab on four consecutive days and did a cognitive test three times every day. Why does JASP use the Durbin Test here? I have no missing data.

Thank you in advance for any help.

EJ

hmm I can't even find the Durbin test. Can you attach a screenshot? I'll forward this to our ANOVA expert.

Cheers,

E.J.

MaximusLuminis

Here you can see the screenshot:

It also says in the JASP Help on Repeated Measures ANOVA:

Nonparametrics

The Friedman test is a non-parametric alternative to the Repeated-Measures ANOVA when there is a complete block design. The Durbin test will automatically be selected when there is an incomplete block design.

JohnnyB

Hi MaximusLuminis,

Thanks for reporting this. It's true that JASP automatically switches to the incomplete version of the Friedman test when it detects some missingness somewhere. Are you in the position to share the data set and JASP file with me, so that I Can look further into this myself? If so, you can email it to j<dot>b<dot>vandoorn <at> uva <dot> nl.

Kind regards,

Johnny

MaximusLuminis

Hi Johnny,

I have sent you an email with the data set and JASP file.

twiggy

Hello,

I'm having the same problem with MaximusLuminis.

I conducted a 3-way repeated measures Anova (nonparametrics), and then it automatically run a durbin-test. However, I don't think my data is "Incomplete Block Design".

Another problem is, when I report the main effect of factors, which one should I use? p or pF??

I'm using the latest version of JASP( 0.14.1 ).

Thank you!

JohnnyB

Hi @twiggy and @MaximusLuminis ,

I have looked into the code that decides whether to use the Durbin test or Friedman test, and it was indeed too permissive. I have updated the code, and it should be fixed in the new JASP release (0.15, hopefully released this week or the next).

Cheers,

Johnny

JohnnyB

Hi @twiggy and @MaximusLuminis ,

I have looked into the code that decides whether to use the Durbin test or Friedman test, and it was indeed too permissive. I have updated the code, and it should be fixed in the new JASP release (0.15, hopefully released this week or the next).

Cheers,

Johnny