Searchlight:: Add the current (center) voxel to its own sphere

[MS Al-Rawi]

Hi ...

I've started using the searchlight function two days ago and I hope to get better results than ANOVA. I was running (using debug mode) mvpa-toolbox searchlight and I noticed that the center voxel was already in the cur_adj_list (line 109 of statmap_searchlight.m). Thus, we might either remove the following lines:

% //  statmap_searchlight.m : lines 113-118

if args.add_center_voxel

    % add the current (center) voxel to its own sphere

    vox_idx = [v cur_adj_list];

  else

    vox_idx = cur_adj_list;

  end

or, reverse the statements to either remove the center or keep it. I must note that I tried different radii values (including zero yeilding two voxels due to above). Without the suggested modification, we'll just have a repeated voxel if we chose to add the center, and the center will always be there for the other choice (If that was the purpose of adding the center then the code should stay the same).

Regards

Al-Rawi

Aveiro - Portugal

I wounder if this still valid (For those who have already know how to use Subversion or who would simply like to browse the source inside your web browser the following link will take you where you need to go: https://www.csbmb.princeton.edu/repos/distpat/trunk/mvpa/ ). 

User authentication was needed!?

[Greg Detre]

> I wounder if this still valid (For those who have already know how to
> use Subversion or who would simply like to browse the source inside
> your web browser the following link will take you where you need to
> go: https://www.csbmb.princeton.edu/repos/distpat/trunk/mvpa/ ). User
> authentication was needed!?

fyi, we're slowly moving hosting to google code, which has a nice
Subversion code browsing interface for exactly this purpose

[Greg Detre]

good catch! thank you.

statmap_searchlight.m's default for add_center_voxel *should* be false.
(in fact this is what the help says it is, but the actual default is
currently set to true). as the help says, ADJ_SPHERE includes it, and so
there's no need to add it a second time.

i've adjusted this in my local version of the code but the SVN is down
right now, so i can't commit the change or wade through the archives to
figure out when this changed.

g