[pymvpa] ERNiftiDataset with different event durations

[Matthias Ekman]

Hi Yaroslav,

Yaroslav Halchenko wrote:
> but then what would you do with such an array? ;) you would have
> different number of features in different samples -- do you want to
> consider it as a "sparse" array may be -- so some samples would not have
> all features present, ie equivalent if we just had them zero-ed out?
>
> what is the real usecae, why don't you like to have them all to have
> maximal duratoin (ie 8sec)?
>   
May be I totally misunderstood the concept of spatio-temporal 
classification,.. :-)

Lets assume an experimental design looking like this:

# Block I

CUE_A
DELAY_A
DELAY_A
TGT_X
--REST

# Block II

CUE_B
DELAY_B
DELAY_B
DELAY_B
TGT_Y
--REST

# Block III
CUE_B
DELAY_B
TGT_Y
--REST
...


Lets assume I want to classify the delaytype (A vs. B). In the first 
case my delay takes 4 sec (TR=2) in the second case the delay covers 
6sec, in the third case 2sec. and so on.
Now you suggest to take the maximum event duration (6sec. in this 
example)? As I understand the Erniftidataset, this would cause the 
problem, that the classifier is also trained on the features belonging 
to the TGT_X class in Block I and TGT_Y + REST class in Block III.

My usecase is, that the Boxcarmapper should only cover the corresponding 
features for one class.

Am I on the woodway :-) ?

Bests,
Matthias


> On Thu, 16 Jul 2009, Matthias Ekman wrote:
>
>   
>> Hi,
>>     
>
>
>   
>> I am currently analysing an event-related design with different
>> event durations (4, 6 or 8 sec.). As far as I can see, it is not
>> possible to use the ERNiftiDataset for this purpose cause it takes
>> the maximum boxlengths for "all" events.
>>     
>
>