> I was just wondering about a comment you made below. You talk about a region
> x condition interaction which you could do by extracting the beta and using
> SPSS or some other stats software.
>
> I was under the impression that you couldn't do this kind of analysis
> because underlying anatomical differences between regions can produce a
> change in BOLD irrespective of any experimental or group difference. So area
> A may be modulated by attention for instance less than area B, not because
> of any real difference in the effects of attention, but because anatomical
> factors mean that the BOLD response has less 'room' to vary.
>
> Is this wrong?
Geoff -
It's neither right nor wrong; I think it's a matter degree.
One could argue that comparisons across regions are potentially
confounded by differences in the mapping from neural to haemodynamic
responses in different regions (owing, for example, to variability in the
vasculature across the brain). Differences in the multiplicative gain of
this neural-to-haemodynamic mapping, in particular, could lead to an
interaction in the haemodynamic data using an additive model like ANOVA,
even if no interaction were present in the underlying neural activity.
Nonetheless, I would still argue that the presence of an additive interaction
is still better evidence than no such interaction. Note that similar issues arise in
possible mappings from psychological processes to behavioural measures
(e.g, whether or not an RT measure of priming, RT1-RT2, scales with absolute RT, RT1).
One really needs a good measurement model (established by independent data).
SPM2 has one in the form of the HDR that is used as the forward model in DCM.
You could "deconvolve" with this first, and put the resulting "neural estimates" into
an ANOVA, if you knew the different heamodynamic parameters of the different regions.
Rik
>> Mauro -
>>
>>> The problem is, as Matthew correctly points out, that people (1) tend
>>> to shorten the statement to "area X is activated by task A" and (2)
>>> may go beyond the data concluding "area X differs from area Y for
>>> task A". Concerning (1), authors should change their habits (make
>>> their constrats really explicit and avoid incomplete statements) and
>>> reviewers shouldn't accept fuzzy statements.
>>
>> Absolutely. More convincing evidence for your claim (2) would be a
>> Region-by-Condition interaction (ie, gets round thresholding effects
>> on the two "simple effects" in regions X and Y).
>>
>> [In fact, I think there are a few more criteria too, which if you are
>> interested,
>> are spelled out in:
>>
>> Henson, R.N.A. (2005). What can functional imaging tell the experimental
>> psychologist? Quarterly Journal of Experimental Psychology, A, 58, 193-233.
>>
>> (let me know if you want a copy)].
>>
>> Such interactions with Region cannot be tested in SPM, which is a mass
>> univariate
>> approach (ie space is itself not "factorised"). It is easily done if you
>> extract the
>> data for a few ROIs. What would appear slightly more complicated is to test
>> such
>> interactions over the whole brain (i.e, how to factorise all voxels). Here
>> multivariate
>> approaches may be more fruitful.
>>
>> Rik
>>
>> ----------------------------------------
>> Dr Richard Henson
>> MRC Cognition & Brain Sciences Unit
>> 15 Chaucer Road
>> Cambridge
>> CB2 2EF, UK
>>
>> Tel: +44 (0)1223 355 294 x522
>> Fax: +44 (0)1223 359 062
>>
>> http://www.mrc-cbu.cam.ac.uk/~rik.henson
>> ----------------------------------------
>>
>
>
|