Print

Print
Hi Donald,

Thanks for your help. Could you explain a little more? I have some questions below:

It's important to model ALL trials in someway or another. Thus, I would recommend that you have 10 conditions, one for each trial. Then at the contrast level, you can compare the 1st and 10th one. With only 1 repetition of each trial, the estimates might not be the best.
At the contrast level - how would I set up the contrast between the 1st and 10th trial if my suspicion is that there is more activation in trial 1 than trial 10 for example.. could I put '1' for trial 1 and '-1' for trial 10 and 0's for everything else?
Importantly, where would I input my behavioural scores in this scenario? It's important to me to be able to relate activation to behavioural performance scores across trials.


The other option would be to use a parametric modulator for trial number. The limitation of this approach is that it assumes a linear increase over all 10 trials. Noise in the estimation of each trial isn't as much of an issue since you are constraining the model to have an increase from trial to trial.
Previously, I inputted my behavioural scores as values of Parametric Modulator (my behavioural scores didn't increase linearly). The idea was that the activation across trials changes with behavioural changes across trials. But, didn't show me any effect.

I think the most interesting is to use the behavioural scores in my model. I am just not sure how, since my previous analysis did not show me an effect. That is why I am thinking to narrow down to comparison to trial 1 and trial 10 (because there should be a big change there even if there isn't from trial to trial).

I appreciate your help.

Thanks,
Joelle


On Sun, Jun 14, 2015 at 10:11 PM, MCLAREN, Donald <[log in to unmask]> wrote:
Joelle,

It's important to model ALL trials in someway or another. Thus, I would recommend that you have 10 conditions, one for each trial. Then at the contrast level, you can compare the 1st and 10th one. With only 1 repetition of each trial, the estimates might not be the best.

The other option would be to use a parametric modulator for trial number. The limitation of this approach is that it assumes a linear increase over all 10 trials. Noise in the estimation of each trial isn't as much of an issue since you are constraining the model to have an increase from trial to trial.

Since you've already done the second way, you can try the first way. The lack of repetition of each trial may make it hard to find significance of a difference over trials, if there is an increase.

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
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, Jun 14, 2015 at 6:24 AM, Joelle Zimmermann <[log in to unmask]> wrote:
Hi - I'm writing with the hopes that somebody can give me advice about how to formulate a particular SPM analysis I want to do (let's start with first-level).

My data:
  • 10 trials (where subjects have fMRI measurement and perform behavioural task). I have a single behavioural performance score per trial for a subject.
  • in between the trials are short 'rest' periods
My goal:
  • Compare first and last trial - to see whether activation changes between first trial and last trial underlie changes in behavioural performance score between first and last trial (there is indeed an increase in behavioural score from first to last trial).
My idea is to set up each of the 10 trials as a 'condition'. Alternatively, set up perhaps only the first trial as a condition, and the last trial as a condition. Where in the design can my behavioural scores for the first trial and the last trial go? 

Additional info:
  • I've previously ran an analysis, setting up one 'condition', with 10 onsets (ie the trial onsets), one Parametric Modulator (with the 10 behavioural scores as 'values'). I set up a t-contrast, where I looked at only the second column (ie my Parametric Modulator) such as 0 1 0 ...
  • I did not get any significant voxels this way.
  • This is why I am thinking of only comparing 1st and last trial (rather than all 10 trials).

Any pointers would be very helpful.
Thanks,
Joelle