The procedure is:

- save residuals by performing a standard ANOVA

- use Aggregate to determine effects for group means (mij for interaction, ai as first factor, bj for second factor)

- eliminate interaction effect from residual to determine interaction (ran), eliminate single effects to determine main effects (ra)

Interaction effect: ek – (αβij – αi – βj + μ)

Main effects: ek – (αi + βj–2μ)

- transform residuals into ranks - run complete model on these ranks for interaction / main effects, respectively

The syntax is:

Code:

```
Unianova x by group drugs
/save=resid (rab)
/design=group drugs group*drugs.
Compute ra=rab.
Aggregate
/outfile=* mode=addvariables
/break=group drugs /mij=mean(x).
Aggregate
/outfile=* mode=addvariables
/break=group /ai=mean(x).
Aggregate
/outfile=* mode=addvariables
/break=drugs /bj=mean(x).
Aggregate
/outfile=* mode=addvariables
/break= /mm=mean(x).
Compute rab=rab-(mij - ai - bj + mm).
Compute ra=ra-(ai + bj - 2*mm).
Rank variables=ra rab (A)
/rank into rar rabr.
Unianova rabr by group drugs
/design=group drugs group*drugs.
Unianova rar by group drugs
/design=group drugs group*drugs.
```

My initial ANOVA syntax would look like this:

Code:

```
Unianova x by condition repetition subject
/random=subject
/method=SSTYPE(3)
/intercept=INCLUDE
/save=RESID (rab)
/criteria=ALPHA(.05)
/design=condition repetition subject condition*repetition condition*subject repetition*subject.
```

This is my first version of the story:

Code:

```
Aggregate
/outfile=* mode=addvariables
/break=condition repetition /mij=mean(T1).
Aggregate
/outfile=* mode=addvariables
/break=condition /ai=mean(T1).
Aggregate
/outfile=* mode=addvariables
/break=repetition /bj=mean(T1).
Aggregate
/outfile=* mode=addvariables
/break=subject /ck=mean(T1).
Aggregate
/outfile=* mode=addvariables
/break= /mm=mean(T1).
Compute rab=rab-(mij - ai - bj + mm).
Compute ra=ra-(ai + bj + ck - 3*mm).
Rank variables=ra rab (A)
/rank into rar rabr.
```

Any tips are greatly appreciated!

Thanks!

(Source: http://www.uni-koeln.de/~a0032/statistik/buch/nonpar-anova-buch.pdf, in German)