Dear Serena,
Luca is correct about the final contrast question.
As for the F-tests, they always tell you when the
combination of one *or* more EVs is significantly
different from zero. So in your first case, zfstat1
tells you when any combination (one or more)
of the EVs is substantially different from zero.
You certainly *do not* need them all to be different
from zero. It is asking the *or* and not the *and*
type of question. So if you get a significant result
from zfstat1 then you don't know which of the
EVs (or which set of EVs) is non-zero - to do this
you need to look at the individual contrasts (C1-C8).
Also, the F-test is always unsigned, which means
that it responds equally to positive *and negative*
changes in the contrasts (unlike the T-tests which only
respond to positive changes).
So zfstat3 will tell you where either C2 *or* C3
*or both* are substantially different from zero.
And that could be because they are positive or
negative - either way, and in any combination.
Finally, zfstat2 is an F-test only on contrast C1
(so I assume you mentioned C2 in your question
by mistake). If you see a stronger result in zfstat2
compared to zstat1 (i.e. in F2 vs C1) then this is
most likely to be caused by negative changes
which C1 will not pick up (the individual T-tests
are signed and so only look for positive changes).
Otherwise they use the same information and the
F-test threshold is slightly higher than the T-test
one since it is two-sided, and so there may be some
things that pass the T-test threshold that do not
pass the F-test threshold.
I hope this answers your questions.
All the best,
Mark
On 10 Jul 2011, at 11:00, sere f wrote:
> Hi All,
>
> I need some advice to understand well the activation maps that I obtain in ER design.
> I use 8 FIR basis functions in Real EV setting.
> This is the setting for contrasts and F-test.
>
> F1 F2 F3
> C1 10000000 * *
> C2 01000000 * *
> C3 00100000 * *
> C4 00010000 *
> C5 00001000 *
> C6 00000100 *
> C7 00000010 *
> C8 00000001 *
>
> I obtain 8 zstat maps, F1, F2 and F3 zfstat maps.
>
> Does F1 zfstat map tell where the activation of all 8 functions is significantly different from 0?
> F2 zfstat map has a bigger activation then C2 zstat map....is it due by the p-value on C2?
> Does F3 zfstat map tell where the activation of C2 and C3 is different from 0?
>
> In addition, the contrast [0,1,1,0,0,0,0,0] asks about the average response of 2 basis functions versus baseline.
> Is the contrast [-0.33,1,1,-0.33, -0.33, -0.33, -0.33, -0.33] different from the previous one?
> And what is the info that they give compared F3 zfstat?
>
> Thanks in advance for any suggestions.
>
> Serena.
|