Question: parametric design encoded with trial_type and amplitude in _events.tsv -- howto prescribe?

We are digging but decided to ask just ask, if we have _events files with onset, duration, trial_type, and amplitude where trial_type is a categorical variable, how should 1st level design be prescribed?

background

Our main confusion is the early on separation of all those additional columns into separate variables which then share only the onset and duration. But in our case, we will have more than a single event for the same onset with different categorical label and having different amplitudes. So we wonder how they would get married again without being mixed up. Example (using spaces here instead of tabs):

onset duration trial_type amplitude
0        0.1         type1       0.2
1        0.1         type2       0.3
0        0.1         type3       0.4
...

so it is important to have later on trial_type.type1 event with amplitude 0.2 and trial_type.type3 event with amplitude 0.4 both happening at onset 0. With separation of the “variables” (now it would be trial_type and amplitude_ … yet to whine about that) it is not clear how to meld them back.

[tyarkoni]

You would do this via transformations. So, for example, if you want a separate regressor for each trial_type value, but with amplitude specified by amplitude, you can use the Split transformation (i.e., split amplitude by trial_type). There are various other transformations you can look at that collectively should allow you to do almost anything you need to (well, within reason—we’re still following the 80/20 rule).

[adswa]

Thx, @adelavega, I’ll keep it in mind!