Dear Hengyi,
I believe most of the information you require can be found in the
manual, but anyway:
> is there any requirement on the least number of subjects?
>
No, you can do ER case studies as well as group studies, similar to
block design experiments. The number of trials for each event type, however,
should be adequate. This has been estimated at 30-40 in previous mails
(e.g.,
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0004&L=spm&D=0&I=-3&P=3485 ),
but depends on a number of factors (stimulus paradigm, contrasts you're
interested in (main effects/interactions), hardware (e.g., field strength)).
> Which contrast should we select? F-contrast or t-contrast? IS there any
> difference between them? If we want to get the spm results of both the
> stim event contrast themselves (compare to null state) and the contrast
> between the two stims, which kind of contrast should we use, respectively?
>
T-contrasts are one-sided tests of the type 'A > B' or 'A>C';
whereas F-contrasts are two-sided ('A differs from B') and can be used for
tests involving > 1 comparisons (e.g., 'A differs from B' and/or 'A differs
from C').
From the above, if you have two conditions, say A and B, and a null
event C, and you specify these in that order (A B C), you could use
A>B: 1 -1 0 (T-contrast)
A>C: 1 0 -1 (T-contrast)
B>C: 0 1 -1 (T-contrast)
and so on. Simple F-contrasts are analogous to T-contrasts, e.g. 'A
differs from B' is 1 -1 0 or -1 1 0, whereas 'A differs from B' and/or 'A
differs from C' would be
1 0 -1
0 1 -1 (two lines).
> Thanks for your comments!
>
Good luck
Dick Veltman
|