Print

Print
Hi again,


also, if I only do one EV as  search RT - 3 column specification- how can FSL know whether each RT is linked to a valid or invalid trial? I guess this is simply because the first column for the onset of the valid and invalid events matches - or should I do 2 EVS instead one for search valid RT and another for search invalid RT - but perhaps this does not make any difference

Yes you definitely need separate EVs for the separate conditions.

just to clarify. If you use one RT modulated EV per event type, yes then you need separate EVs. If you use one unmodulated EV per event type and then one modulated to model RT then the latter should be common to both event types.

Having said that, thinking a little bit more about it I no longer think the 3 EV model is a good idea. My thought was to somehow model the within-type variability caused by RT, while leaving any between type variability to the comparison of the 2 even types. But I now realize that one can express the 2 EV case as a linear combination of the 3 EVs, so I don't really think that it would buy you anything.


lastly, how about the  memory RTs? Can I just ignore these in the model specification. My main aim s to compare Valid>Invalid activity and this is not related to the memory RT but may only be related to the search RT---on the other hand, it is true after the search response, there followed a memory response, and perhaps this should be included in the model?

The tricky thing about your kind of analysis (complex compound tasks) is to make sure the different types of events differ as little as possible, ideally only w.r.t. on of the sub-events. If you have good reason to believe that the "memory response" is identical for valid and invalid clues, then yes model it as a separate EV (that is common to both event types). If you have reason to believe they might be different then you should not model it with an extra EV, instead allowing the variance to go into the event specific EVs.

Does that make sense?

Good luck Jesper





hope this is not much of a hassle;

best wishes!






---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------