Print

Print


Hi, Anderson

I tried running the CCA in PALM using v108 but it complained about line
saying "Reference to a cleared variable o"
M = plm.Pset{p}*plm.Rz{y}{m}{c}{o}*plm.X{y}{m}{c}{o};

So I modified the line to:
M = plm.Pset{p}*cell2mat(plm.Rz{y}{m}{c})*plm.X{y}{m}{c};

and it is running fine.The reason I had to convert cell 2 mat as cell Matlab doesn't support cell multiplication.

The script did not finish running as yet but it has produced *_vox_cca_c1 and *_vox_cca_c2. These files are identical although c1 = 1 0 0 ... and c2 = -1 0 0 ...

Is this the expected behaviour? and whether the changes that I did in those lines are correct?

Thanks

Regards

Virendra

On Wed, Aug 9, 2017 at 7:04 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Virendra,

Just one call. It won't change what you'd like to test: after removing EV2, and the mean-centering that is done internally when CCA is used, the EV1 will be coding group differences. The contrasts should then be:

C1: [1 0 0 ...]
C2: [-1 0 0 ...]

That's it!

All the best,

Anderson


On 9 August 2017 at 10:40, neuroimage analyst <[log in to unmask]> wrote:
Thank you Anderson

So, should I then run CCA for the 2 groups separately in 2 different call to palm? How to set the contrast for 2 groups?

Thanks 

Regards 

Virendra 


On Aug 9, 2017 6:23 AM, "Anderson M. Winkler" <[log in to unmask]> wrote:
Hi Virendra,

Thanks for sending the files and for the patience.

This error can be fixed by replacing, in line 2047 of palm_core.m, where it reads:

for t = find(yselq),

for the following:

for t = find(yselq)',

That is, an apostrophe was missing, which caused the variable t being not a row but a column, which on its turn changes the behaviour of the for-loop.

In looking into this, it seems that there was also an issue with the intersection mask across modalities, which not being produced; this has now been fixed elsewhere in the code (it wouldn't affect your analysis, but could crash with different data). This has now been fixed too, in alpha108.

A separate comment on your data: the design has one column for each of the two groups. This is generally fine, except for CCA, where we mean-center all columns, and then the design becomes rank deficient. To get around it, and avoid warning messages, just remove the 2nd column of your design matrix and also the corresponding bits of the contrasts.

Hope this solves the problem. If not, please let me know. Thanks!

All the best,

Anderson


On 6 August 2017 at 22:30, neuroimage analyst <[log in to unmask]> wrote:
Hi, Anderson.

Thank you for looking into this. You should have received a dropbox link with all the relevant files in it. Please let me know if anything else is needed from me. I really appreciate your time and help.

Regards

Virendra

On Sun, Aug 6, 2017 at 4:03 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Virendra,

Could you send me your inputs, design files, etc? You can put in some online service like Dropbox and send me the link off-list then.

Thanks, and apologies for the persistent bug...

All the best,

Anderson


On 6 August 2017 at 11:28, neuroimage analyst <[log in to unmask]> wrote:
Thanks, Anderson.

However, it throws the following error now.

Undefined function 'qr' for input arguments of type 'double' and attributes
'full 3d real'.

Error in palm_core>cca (line 3197)
[Qy,~]  = qr(Y,0);

Error in palm_core (line 2048)
                                Q{m}{c}(t) =
                                cca(plm.Yq{m}{c}(:,:,t),M,opts.ccaparm);

Error in palm (line 81)
palm_core(varargin{:});

I would greatly appreciate your response.

Thanks

Regards

V

On Sun, Aug 6, 2017 at 5:01 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Virendra,

Thanks for letting me know. This should now be fixed in version alpha107. If the problem persists (or some other appears!) please feel free to ask again.

All the best,

Anderson


On 5 August 2017 at 01:43, neuroimage analyst <[log in to unmask]> wrote:
Hi, Anderson

I am trying to run PALM CCA for 3 modalities with the following line:

palm -i mod1.nii.gz -i mod2.nii.gz -i mod3.nii.gz -m mask.nii -d design.mat -t contrast.con -o results -npcmod -npccon -mv CCA -T -logp -accel tail -quiet -noniiclass -nouncorrected -savedof -saveparametric -Tnpc -demean -n 100

However, it throws an error:
Undefined operator '*' for input argument of type 'cell'
Error in palm_core (line 2044)
M=plm.Pset(p)*plm.Rz{m}{c}*plm.X{m}{c}
Error in palm (line 81)
palm_core(varargin(:))

I would greatly appreciate if you could please help me resolve this error.

Thanks

Regards

Virendra