We are analyzing data from a bloc design fMRI study (2 condtitions Right eye (R) and Left eye (L), two different groups of subjects Young (Y) and Old (O). We want to perform group comparisons on the second level. We are interested in the four contrasts:
(YR-YL)>(OR-OL)
(YL-YR)>(OL-OR)
(OR-OL)>(YR-YL)
(OL-OR)>(YL-YR)
Contrasts at the first level were computed as follows:
Con_0003.img was computed as R-L for both groups using an implicit baseline contrast specification [1 0 0]
Con_0004.img was computed as L-R for both groups using an implicit baseline contrast specification [0 1 0]
For the second level group comparisons, we select ten Con_0003.img files for the young subjects and then ten Con_0003.img files for the old subjects.
Within a two sample t-test we specify the "group? (2)" input as 11111111112222222222
To compute the (YR-YL)>(OR-OL) contrast we specify [1 -1 0] and to compute the (YR-YL)>(OR-OL) contrast we specify [-1 1 0]
In a subsequent two sample t-test, we select ten Con_0004.img files for the young subjects and then ten Con_0004.img files for the old subjects.
Again we specify the "group? (2)" input as 11111111112222222222
To compute the (OR-OL)>(YR-YL) contrast we specify [1 -1 0] and to compute the (OL-OR)>(YL-YR) contrast we specify [-1 1 0]
It turns out that the contrasts are redundant:
(YR-YL)>(OR-OL) is identical to (OL-OR)>(YL-YR)
and
(YL-YR)>(OL-OR) is identical to (OR-OL)>(YR-YL)
In the second level analysis, should we have specified our contrasts as [1 0 0] and [0 1 0] rather than [1 -1 0] and [-1 1 0]?
Of so, why?
Jerry Allison
|