Hi Nan,
What you describe is fine for making sure that all responses (A through D) are positive, as long as you are using the option (Z>0) for the contrast masking.
There is still one thing that it doesn't distinguish though.
Imagine that C7 has a positive response, then there are still two possible modes that this can happen (even when C1-C4 are all positive). These are:
(i) A>B is positive and bigger than C>D (which could be positive, zero or negative)
(ii) A>B is negative (or zero), but C>D is even more negative
That is, C7 could be driven by a situation where B>A but D>>C.
It might be that you don't care about the above distinction, which is fine and in that case you have done everything that you need to have done. If you did want to pick apart the results further, then you can distinguish between case (i) and case (ii) above by also including C5 in the contrast masking, as this will restrict your results to just case (i). A similar thing can be done for C8 of course.
I hope this helps.
All the best,
Mark
On 3 Jul 2014, at 15:25, Nan Wise <[log in to unmask]> wrote:
> Dear Experts,
>
> Dissertation defense is looming and I have yet another question about contrast masking.
>
> If I am bringing up two differential contrasts from my first levels:
>
> A>B
> C>D
>
> having also run these contrasts against baseline on the first level
> A>0
> B>0
> C>0
> D>0
>
> And am interested in doing a comparison between those two differential contrasts (A>B ) > (C>D) on the second level to see what the difference between them is, while making sure that what shows up is indeed driven by the positive activations rather than deactivations:
>
> Is it valid to do higher level analysis to compare (A>B ) > (C>D) as outlined below?
>
> EV 1- A>0 (baseline)
> EV2- B> 0
> EV3- C>0
> EV4 - D>0
> EV 5 A>B
> EV6 C>D
>
> And then do the following 8 contrasts
>
>
>
> EV1 EV2 EV3 EV4 EV 5 EV6
>
> C 1- A>0 Group mean 1 0 0 0 0 0
>
> C2- B> 0 Group mean 0 1 0 0 0 0
>
> C3- C>0 Group mean 0 0 1 0 0 0
>
> C4 - D>0 Group mean 0 0 0 1 0 0
>
> C 5 A>B Group mean 0 0 0 0 1 0
>
> C6 C>D Group mean 0 0 0 0 0 1
>
> C7 ( A>B ) >( C>D) 0 0 0 0 1 -1
>
> C8 (C>D) > (A>B) 0 0 0 0 -1 1
>
> And then doing the contrast masking of both C7 and C8 with C1, C2, C3, and C4?
>
>
> Is this at all valid? What I am trying to accomplish is to determine the differences between the two differential contrasts in both directions.
>
> Thanks so much.
>
> The very weary Nan Wise
|