Dear Bradley,
Set-up 2 is what you want, as set-up 1 is unable to model the
situation where the different conditions produce different
response amplitudes, and so all the unmodelled difference
would end up being treated as residual noise - which is not
what you want.
The contrast [1 1] and [0.5 0.5] are completely equivalent.
You can multiply any contrast by a constant and it is unchanged
statistically.
Note that the [1 1] contrast asks about the average response
versus baseline. You can also get an F-test across the [1 0]
and the [0 1] contrasts and this will ask the question where
you get a response to either or both conditions.
All the best,
Mark
On 9 Jul 2011, at 17:56, Bradley Sumner wrote:
> Would the following two feat set-ups be equivalent/correct?
>
> Two conditions, A and B (for example two image types, scenes-A and faces-B)
>
> Set-up 1: three column text file with onsets for condition A trials and condition B trials (covering all trials as if A and B were both part of one condition C) as a single EV.
>
> Contrast [1] for all trials > baseline
>
> Set-up 2: three column text file with onsets for trials in condition A entered for EV1, three column text file with onsets for trials in condition B entered as EV2.
>
> Contrast [1 1] for all trials > baseline
>
> And a related question, would the contrast [1 1] in set up 2 be equivalent to the contrast [.5 .5]?
>
> Thank you.
>
|