Dear Sina,
I meant an ANOVA (within-subjects), but having reread your design, I think
you could also us a paired t-test, because you only have 2 conditions. (An
ANOVA (within-subjects is an expansion of the paired t-test for >= 3
conditions.)
From you description I assume that your con-image encode the effect of
each condition, and that you want to compare these, e.g. in a differential
t-contrast at the second level. Right?
If this is the case, you could use a paired t-test, resulting in a design
matrix like this:
1 0 1 (subj1)
0 1 1
1 0 1 (subj2)
0 1 1
...
You would then choose "non-sphericity correction - yes", and "replication
are over subj/repl (20)". The combination of these correct for violation of
the i.i.d. assumption (identical and independent errors). The basic
non-sphericity correction will correct for non-identical error, the
"replication are over" will correct for dependent errors.
As contrast weights, you would enter:
1 -1, tests for cond1 > cond2
-1 1, tests for cond1 < cond2
Now, you mentioned something about a conjunction, but I didn't understand
that part. What are the two contrasts that you want to combine in a
conjunction?
Cheers,
Jan
Sina Alexa Trautmann wrote:
> Dear Jan,
> I am sorry to bother you again, but I was wondering what exactly you
> meant by "within-subjects" ANOVA. Did you simply mean: ONE-WAY ANOVA
> with spherecity correction and the repeated measure ("yes")-button (this
> is what I did now) _*or*_ did you mean the ANOVA (within-subjects)
> implemented in SPM2. Because as far as I understood, it is not possible
> to do a conjunction with the ANOVA (within-subjects) because the factors
> need to be "independent". Besides it is also possible to do the one-way
> ANOVA with my below described design because we are already looking at
> second level contrast. Alexa wrote this on the list recently, so I guess
> my approach is ok?!
> Greets, Sina
>
> Jan Gläscher wrote:
>
>> Dear Sina,
>>
>> "replication are over ..." refers to the random effect in your design,
>> the independent replications of your experiment, i.e. the subjects.
>> However, you might want to consider a within-subjects (=repeated
>> measures) ANOVA to account for the two runs that you acquired per
>> subject. In this design the replication are also over subjects/repl.
>>
>> Cheers,
>> Jan
>>
>>
>>
>> Sina Alexa Trautmann wrote:
>>
>>> Dear SPMers!
>>>
>>> I got very confused reading the adds of the list about the
>>> non-sphericity correction in SPM2. I have the following problem:
>>> I want to do a one-way ANOVA in SPM2 with 20 subjects which each
>>> completed 2 runs meaning I have 2 conditions (1 condition per run).
>>> For each condition I have 1 contrast per subject (meaning 2x20
>>> contrasts). If I want to do a non-sphericity correction, I would
>>> choose "replications are over: group(2)" because the conditions are
>>> measured in one session one after the other. Is that correct or do I
>>> have to choose "repl(20)"?
>>>
>>> I am looking forward to your advice and thank you in advance!
>>>
>>> Greets, Sina Trautmann
>>>
>>>
>>
>
--
Jan Gläscher, Ph.D. Caltech Brain Imaging Center
+1 (626) 395-4976 (office) Caltech, Broad Center, M/C 114-96
+1 (626) 395-2000 (fax) 1200 California Blvd
[log in to unmask] Pasadena, CA 91125
|