Vadim -
> Specifically:
> 1) Is more blocks with shorter length (for example, block length 6s =
> 3 TRs) better than less blocks with longer length (for example, block
> length 12s = 6 TRs)?
For two conditions, alternating blocks of 12-15s would be best, since
the fundamental of the power for a [1 -1] contrast between those
conditions is close to the dominant frequency of the HRF (assuming you
model those blocks as epochs, eg trials within blocks occur frequently
enough). For more than two conditions, the same logic applies, though
contrasts comparing a subset of those conditions will become less
sensitive, because further apart in time on average (and close to 1/f
noise levels if too many conditions/too long blocks). A repeated block
order (A,B,C..., A,B,C) will minimise this problem (compared to a fully
randomised ordering of conditions across the experiment, eg
A,B,B,C,A,A,A,C...), though may introduce order confounds (eg
psychological effects); a permuted design (e.g, A,B,C..., B,C,A...,
C,A,B..., B,AC...etc) might be a reasonable compromise.
> 2) What is the minimal between block fixation length should be in
> order that the blocks would not be contaminated? I know, that some
> labs use functional localizer block designs without any between block
> fixation interval at all.
You don't need fixation or "rest" conditions between blocks. There is no
need (in a linear system) for the HRF to "return to baseline". There is
no absolute baseline for all brain regions anyway. You only need to
include fixation blocks if fixation is actually a condition that
interests you (ie meaningful for a specific brain region).
> 3) If between block interval is short and the signal doesn't return to
> baseline, does it make sense to to make variable between block
> interval (type of jittering) ?
See above. No need to jitter with blocks (with epoch-models) from point
of view of HRF estimation (though small jitter may help to avoid
aliasing of oscillatory noise, though unlikely to be phase-locked).
> 4) Does by more efficient design term we understand that the design
> would be more efficient for any type of analysis (GLM, raw data
> pattern classification etc) or it is analysis specific (GLM)?
More efficient for a *specific contrast* in a *linear* model.
Hope this helps; if not, you could look at the SPM course slides on the
website about efficient design.
Rik
--
-------------------------------------------------------
Dr Richard Henson
MRC Cognition & Brain Sciences Unit
15 Chaucer Road
Cambridge
CB2 7EF, UK
Office: +44 (0)1223 355 294 x522
Mob: +44 (0)794 1377 345
Fax: +44 (0)1223 359 062
http://www.mrc-cbu.cam.ac.uk/people/rik.henson/personal
-------------------------------------------------------
|