time-course / % change

Hello All,

I've run into something I can't seem to figure out, even with the FAQ 
and the mailing list.  Perhaps I missed something.  In any case, I hope 
these are simple enough questions to be answered easily.

- We used MarsBaR to extract % signal change and time courses (fitted 
event) from several ROIs within out data.  When we look at the % change 
numbers and the time course numbers, they don't match up.

- Shouldn't the peak (abs max) time course be our peak % change too?

- In an e-mail reply on the mailing list, it was suggested that perhaps 
the time course is calculated with one HRF and sometimes the percent 
change is calculated with another (e.g. from the SPM estimate).  Could 
this be the reason why our % change and time course peaks aren't the same?

- Or, are the differences in these numbers due to the 'SPM units'?  Is 
it that one of these measures (time course or % change) is in 'SPM 
units' and the other is not?  If so, how can we convert one into the other?

- We also have been trying to figure out how to calculate % change/time 
courses manually but we can't seem to find the variable that contains 
the time information.  We were thinking that if we could find that, and 
had the betas (which we have located), we would be able to manually 
calculate the % change and time course.

Really we're looking for a more transparent calculation so we can 
reconcile the different numbers.  I've looked extensively at the FAQ and 
the mailing list but even with that information, I don't feel I have a 

a problem with 'get_marsy'

Hi,

I got an error message as follows when I tryed to extract
time course.

??? Undefined function or method 'get_marsy' for input arguments
of type 'mardo_5'.

The following is the script.

D  = mardo(spm_name);
R  = maroi(roi_file);
Y  = get_marsy(R, D, 'mean');

How can I solve this problem?
I am using matlab 7.6.0.

Thank you in advance!

Best regards,

Michiru

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: time-course / % change

Hi,

> - We used MarsBaR to extract % signal change and time courses (fitted
> event) from several ROIs within out data.  When we look at the % change
> numbers and the time course numbers, they don't match up.

First - I am sure you found this, but there's an item in the FAQ
explaining how the % signal change is calculated.  After that, you
will need the code itself, which is in

http://marsbar.sourceforge.net/doc-devel/latest/marsbar/ <at> mardo/event_signal.html

> - Shouldn't the peak (abs max) time course be our peak % change too?

No, the fitted events use the usual SPM scaling units.  They will be a
bit similar to % signal change, but not the same.

> - In an e-mail reply on the mailing list, it was suggested that perhaps
> the time course is calculated with one HRF and sometimes the percent
> change is calculated with another (e.g. from the SPM estimate).  Could
> this be the reason why our % change and time course peaks aren't the same?

No, I don't think so.  The model of the time course should be the same
for the two calculations - in fact the % signal change uses the fitted
event calculation.

> - Or, are the differences in these numbers due to the 'SPM units'?  Is
> it that one of these measures (time course or % change) is in 'SPM
> units' and the other is not?

Re: a problem with 'get_marsy'

Hi,

> ??? Undefined function or method 'get_marsy' for input arguments
> of type 'mardo_5'.
>
>
> The following is the script.
>
> D  = mardo(spm_name);
> R  = maroi(roi_file);
> Y  = get_marsy(R, D, 'mean');

That's very strange.  get_marsy is a method of the 'maroi' class, so
it should be called as a result of the 'R' object in your script.
What type of thing (>> whos) is 'R' in your script, after you've run
it?

Best,

Matthew

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/

residual time series after FIR model fitting

Hi,

I like to obtain residual time series after FIR model fitting
to calculate functional connectivity.
Could you tell me how I can do this by batch? 

Thank you in advance!

Michiru

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: residual time series after FIR model fitting

Hi,

> I like to obtain residual time series after FIR model fitting
> to calculate functional connectivity.
> Could you tell me how I can do this by batch?

As for SPM, this is not straightforward.  The data and the design are
filtered before estimation, so you will have to replicate this, to get
the filtered residuals (these are the residuals from the model SPM and
marsbar uses).   This involves

Run the estimation with FIR modeling of the events

D = mardo('SPM.mat');
mY = get_marsy(R, D, 'mean');
E = estimate(D, mY)

Filter the time course with the design filter -> fY

Y = summary_data(mY);
fY = apply_filter(E, Y);

Get the inverted filtered design matrix (ifX = pinv(fX), where fX is
the filtered design matrix) from the SPM structure:

SPM = des_struct(E);
ifX = SPM.xX.pKX;

Get the betas from the original estimation:

re siduals using MarsBar and SPMd

Hi Matthew,

I extracted residuals from ROIs using Marsbar and SPMd to see the
difference. They are different. 
I wander if I'm doing anything wrong in Marsbar?

1. design matrix contains confounds to be removed and an empty regressor
2. in Marsbar:
  extract mean timcourses for ROIs ("Extract ROI data");
  estimate results ("Estimate Results");
  obtaining timecourse for the residuals: "MarsBar SPM Graph"; any contrast
for empty regressor; "adjusted"; looking at variable y in the workspace

Obtaining residuals in SPMd is pretty straigtforward: "Scan Details" in
SPMd; extracting mean timecourses for the ROIs using "Extract ROI data" in
MarsBar.

Could you please help?

Thanks,
Roman
--

-- 
View this message in context: http://www.nabble.com/residuals-using-MarsBar-and-SPMd-tp20517526p20517526.html
Sent from the MarsBaR mailing list archive at Nabble.com.

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world

MarsBar batch to get residuals

Hi Matthew,

I've seen the script you wrote in response to a question from Michiru.
Tried to adopt for what I need (extraction of residuals from an ROI). The
result is different from what I get using MarsBar GUI. Could you please
advise what is wrong here.

D = mardo('SPM.mat');
mY = get_marsy(R, D, 'mean');
E = estimate(D, mY)

Y = summary_data(mY);

SPM = des_struct(E);
ifX = SPM.xX.pKX;

B = betas(E);

fitted = ifX * B;
resid = Y - fitted;

Thanks in advance,

Roman
--

-- 
View this message in context: http://www.nabble.com/MarsBar-batch-to-get-residuals-tp20517874p20517874.html
Sent from the MarsBaR mailing list archive at Nabble.com.

-------------------------------------------------------------------------

Re: MarsBar batch to get residuals

I think the next is more likely to be correct:

D = mardo(spm_get); % specify SPM.mat
mY = get_marsy(R, D, 'mean');
E = estimate(D, mY)
residuals = residuals(E);

the timecourses ectracted using this script look very similar to the ones I
get using SPMd but not the same (e.g. correlation coefficient is around
0.95-0.97).
I don't understand what bit I'm missing as both toolboxes seem to be using
SPM mashinery to calculate the residuals ...
would be realy great to learn (even suggestions would be great)

Thanks,

Roman

RRoman wrote:
> 
> Hi Matthew,
> 
> I've seen the script you wrote in response to a question from Michiru.
> Tried to adopt for what I need (extraction of residuals from an ROI). The
> result is different from what I get using MarsBar GUI. Could you please
> advise what is wrong here.
> 
> D = mardo('SPM.mat');
> mY = get_marsy(R, D, 'mean');

ROI creation

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/

_______________________________________________
Marsbar-users mailing list
Marsbar-users <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/marsbar-users