Thank you for the timing files, which clarify what the
2:1 ratio refers to.
You are comparing regressors within a condition, while I
have been trying to reason why they would differ across
conditions. You did say the first regressor was different,
but I still assumed it was classes. Sorry for that confusion.
What you are seeing is simply that the endpoint tents
occupy basically half the number of TRs that the middle
tents occupy. This happens in any case where the events
(or tent centers) are not TR-locked. That is because a
TR that is outside the range of the basis function has no
contribution to its evaluation. The middle tent functions
are complete tents, while the end ones are 1-sided.
This is not something strange about your timing, but
applies to any random timing with tent functions.
On a separate note, TENT(0,15,6) is designed for a
TR of 3.0, not 2.5. I believe you probably want 7 tents,
to make them TR-locked, TENT(0,15,7).
- rick