Hello, experts.
I am looking for theoretical background of using Fourier set to
analyze data from sparse sampling technique.
According to the archives written by Dr. Friston, Dr. Henson, and
other experts, I am supposed to adopt the setting I wrote at the end
of this mail.
Although the method showed some activations, I am not sure of using
Fourier set as basic function.
In sparse sampling technique, we can get only one set of scan per a
stimuli. In addition, SOA is usually not jittered. I wonder whether
the Fourier set picks up false activations.
Do you have any theoretical background of the adoption of Fourier set
and the way written below?
(Any advice on the way would be also appreciated.)
P.S. I tried using one sample t-test to analyze the data acquired in
sparse sampling, but the result was far poorer than that in the above-
mentioned method. Please tell me the reason, if you know why there is
so much difference.
Thank-you for your valuable time.
T.W.
-----------------------------------------------------------------
The 1st level analysis
i) Pretend TR is 3 sec., not real long TR (e.g. 14 sec.)
ii) Never convolve hrf. Instead, use the Fourier set (without
Hanning) or 'mean and exponential decay' as basic function.
iii) Create 7 con*.img files. ( the number of files is
determined by the number of order. I chose the Fourier set with order
3. I had two conditions: A and B.)
e.g. [1 0 0 0 0 0 0 -1 0 0 0 0 0 0], [0 1 0 0 0 0 0 0 -1 0 0 0 0
0] , ...
The 2nd level analysis
iv) Gather each con*. files across subjects. (So I got 7 folders
each of which contained the same type of con* files)
v) Analyze those data by using one way ANOVA.
vi) See the result with F contrast.
------------------------------------------------------------------
---------------------------------------------------------------
Takamitsu WATANABE
The Faculty of Medicine
The University of Tokyo
[log in to unmask]
---------------------------------------------------------------
