Results may be inaccurate. RCOND = 1.034803e-027

Hi everybody,

When I'm regressing out covariates Matlab returns the following to me:

Warning: Matrix is close to singular or badly scaled.
Results may be inaccurate. RCOND = 1.034803e-027.
> In rest_RegressOutCovariates>Brain4D_RegressOutCovariables at 92
In rest_RegressOutCovariates at 54
In rest_RegressOutCovariates_gui>Run_Callback at 238
In gui_mainfcn at 96
In rest_RegressOutCovariates_gui at 26

What can I do about that?

Thanks a lot,
Marisa

Re

Hi!
How many covariates have you set? And how many time points do you have?
This situation happens when you have too few time points but too many covariates. Then almost all the useful informaion is regressed out as covariates.
Best wishes!

covariates

Hi,
My covariates are brain and white mask, CSF and motion-parameters, which should be regressed out. And I have 142 volumes (originally I had 150 volumes, but with 150 I always got the "out of memory" error by matlab, and with 142 it worked properly).
Is there anything I can do about the covariates and the ran out of memory error?
Cheers

fixed with 150 volumes

Hi again,
I've fixed the 150 volume-problem now by using a linux 64-bit machine. The other problem remains unsolved...
Regards,
Marisa

Re

Hi!
Regarding to the error of out of memory, please refer to http://www.restfmri.net/forum/Course#comment-168 , you can turn on the 3GB switch of Windows.

As for the warning of "Results may be inaccurate. RCOND = 1.034803e-027", could you check the covarites which may stored under FunImgNormalizedXXXXX_Covs\xxxx.txt. Please check if the covariates are almost zero, there may be something wrong in this case.

Hi, numbers in my individual

Hi,

numbers in my individual Cov.txt files of one person look more or less like this:

0.0000000000000000e+00 0.0000000000000000e+00 7.9214351171610326e+02 9.1221958797016146e+02 8.9997856529579963e+02
1.0547152000000001e-02 7.2709502000000001e-03 -1.3676035000000000e-02 6.2034036000000001e-04 1.6738022000000000e-04 -2.3931180999999999e-04 7.9038008615149693e+02 9.1149857053999403e+02 8.9576808252114279e+02
-5.4496309000000000e-02 -1.0728404000000000e-01 8.6074365000000000e-02 -3.5257385999999998e-03 -1.9205508000000001e-03 -8.7843722999999998e-04 7.8894800972296821e+02 9.1121836169634139e+02 8.9257379449187590e+02
-5.9235959999999997e-02 -9.3484816999999998e-02 1.5442724999999999e-02 -2.3419474000000002e-03 -2.0580742000000001e-03 -1.5867343999999999e-03 7.8805182695966721e+02 9.1096741286279507e+02 8.9154997221523911e+02
-2.2137424999999999e-02 -2.9264557000000000e-02 1.7055569000000000e-02 -1.6929338000000000e-03 -1.4255422000000000e-03 -1.0443792000000001e-03 7.8753766219904912e+02 9.1047034212430083e+02 8.9181764603920112e+02 and so on...

which, yes, is close to zero. This brings me to my next question: what can I do against this? Don't I have to regress out Whole Brain, White matter, Csf and motion? If yes, how can I do without regressing out my signal of interest?

Thanks again!

Re

Hi!
Could you upload one Cov.txt file here?
It seems that there is something wrong with your data which makes the Cov Matrix close to singular.

Cov.txt

What I did with my data was: realignment, normalizing, temporal filtering (0.009-0.08 Hz, TR=3, 150 Volumes), smoothing, extracting ROI time course (whole brain, white matter, csf), creating the Covs with the code:

RPcov=load('rp_rest.txt');
BCWcov=load('any_person_filtered_smoothed_ROITimeCourses.txt');
Cov=[RPcov, BCWcov];
save('Cov.txt','Cov','-ASCII','-DOUBLE','-TABS');

creating the CovList.txt, regressing out the Covs (whole brain, WM, csf, and motion), creating the mask for seed region and then I ran Func-Con.

Unfortunately I don't know how to attach a file on this mask. I tried to copy/paste it in here but it doesn't appear approriately. Can I send it to you via mail?
Thanks again!

Re

Hi!

For the steps you mentioned, you can also use DPARSF to do it automatically in case there is any mistake.

You can send your Cov.txt to ycg.yan#gmail.com  (@ instead of #)

covariates

Hi,
My covariates are brain and white mask, CSF and motion-parameters, which should be regressed out. And I have 142 volumes (originally I had 150 volumes, but with 150 I always got the "out of memory" error by matlab, and with 142 it worked properly).
Is there anything I can do about the covariates and the ran out of memory error?
Cheers

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.