Hi Franziska,
thank you very much for your response. I am not sure if i get
you right. You modeled both the conditions (2back vs. 0back as
blocks) and the events (every single presented picture in
accordance to the condition), right? This would result in 4
predictors. In such a model the information of the blocks and
events are overlapping each other. However, i checked it out but
unfortunately with the same result: very strange looking design
matrix (nearly all vectors are displayed in grey, no differences
between the predictors from different conditions).
In another approach of analysis I have splitted my long 72
seconds blocks in three smaller 24 second blocks. Now it seems
that my design matrix looks like it supposed to. Due to the
splitted blocks the amount of predictors increased from 2 to 12
(2x 2back + 2x 0back -> 2x3 2back + 2x3 0back). Do you think
one can do that this way? I will check this model with more
subjects in the next days...
@Chris: thank you for your suggestions. I will model my data as
an event related design with respect to the performance of the
subjects in the next week. However, I am afraid concerning low
SNR because of the very few hit/miss/false alarm events...but
i´ll see
thanks and kind regards
christian
Am 1/26/2013 12:20 AM, schrieb Franziska Korb:
[log in to unmask]" type="cite">Hi
Christian,
in a similar 2-back study, we modeled both the events and the
blocks (as a so-called mixed design) - that way you use the HPF
for the events (SPM default of 128sec), but also model
information about the different epochs in your study. When
building the contrasts, you can then "ignore" the events by
giving them a "0" weight and contrast the 2-back vs. n-back
blocks (1/-1). It worked well for us.
Good luck,
Franziska
Am 1/25/2013 9:31 PM, schrieb Chris Watson:
