I personally wouldn't worry about using an implicit mask, but just
plot the data. When you use a mask, your imposing another constraint
on the data. Here are a few examples:
(1) Interaction, mask A2>A1 and mask B2>B1 --> Both must have the same
direction of change, lines are both positives; not a requirement for
an interaction
(2) Interaction, mask A2>A1 --> Only A must have a positive value, B
could be positive, 0 or negative. Less restrictive, but still not a
requirement
(3) Interaction, mask B2<B1 --> B must have a negative value AND A
must have a positive value (A could be negative, but then it wouldn't
show up in the interaction contrast unless the interaction was an F);
again not a requirement for an interaction.
(4) Interaction, mask A2>A1 and mask B2<B1 --> A and B must have
opposite direction of changes, one line is positive and one line is
negative; not a requirement for an interaction
(5) masks -- what is the threshold. you could have neither A or B
being different than 0 and still get an interaction.
(6) This leads to the following approach: Mask the interaction with
A2>A1, but at a very very liberal threshold of .5. This will give you
all the regions with an interaction where A is greater than B. You can
do the opposite to find the other direction with a mask B1>B2.
Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General
Hospital and Harvard Medical School
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain
PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED
and which is intended only for the use of the individual or entity
named above. If the reader of the e-mail is not the intended recipient
or the employee or agent responsible for delivering it to the intended
recipient, you are hereby notified that you are in possession of
confidential and privileged information. Any unauthorized use,
disclosure, copying or the taking of any action in reliance on the
contents of this information is strictly prohibited and may be
unlawful. If you have received this e-mail unintentionally, please
immediately notify the sender via telephone at (773) 406-2464 or
email.
On Sun, Jan 23, 2011 at 8:16 AM, Michael_Mouthon
<[log in to unmask]> wrote:
> Good morning to all SPM community,
> I need to know if the mathematical sign of betas should be taken account during implicit masking of interaction contrast.
>
> Reminder of the standard procedure:
> To disambiguate an interaction contrast with 2 factor and 2 level ((A2>A1) > (B2>B1) and (B1>B2) > (A1>A2) which have both the contrast vector -1 1 1 -1), we usually use an inclusive masking with (A2>A1) and (B2>B1) if we are interest by the comparison ((A2>A1) > (B2>B1).
>
> Now come my question:
> I was surprise to found in the article Sass, K. et al (2009) [Taxonomic and thematic categories: Neural correlates of categorization in an auditory-to-visual priming task using fmri] a different procedure:
>
> They studied the interaction [thematical related>unrelated]>[taxonomical related>unrelated] masking this [thematical related>unrelated] and !!![taxonomical unrelated>related]!!!
> In respect to previous notation, It correspond to performed an interaction ((A2>A1) > (B2>B1) with (A2>A1) and !!!(B1>B2)!!! instead of (B2>B1) as mask.
>
> Why the second mask was inverted in respect to the related>unrelated (2>1) of interest?
>
> In attachment, you will found the contrast estimate plot of this article. We can notice that the comparison [thematical related>unrelated] give a positive difference and that [taxonomical related>unrelated] gave a negative difference.
> The inversion of the second mask [taxonomical unrelated>related] permit to consider a positive difference.
>
> Is this procedure right ? Why ?
> Do you need to take account the mathematical sign of betas instead of the related>unrelated ?
> If the response is yes, how can you manage the pattern differences in the whole brain (this graphic represent only one voxel)?
>
> Thank you in advanced and best regards.
>
>
>
|