Print

Print
Dear all,

Thanks for your advice! One important question I have is: should I add a data point for time bin=0 so that I can show the beta at t=0? Because my stimuli are ON for 6 seconds (ie at time bins1&2) then OFF for 15 seconds. Now the plot looks as if beta is already at max right at the beginning...

I wish to reply to Donald here:
(1) These FIR plots are averaged from 25 subjects.
(2) Yes, I made all trials to begin at the same time relative to the beginning of the TR.
(3) All trials are of the same duration.
(4) If each trial is preferred to be 20-30 seconds, does it mean event-related designs are discouraged from using FIR?

And to Colin, would you share with me more by explaining what you mean by noisy HRFs?

Thanks a lot.
Andy


On Tue, May 6, 2014 at 1:34 AM, MCLAREN, Donald <[log in to unmask]> wrote:
A few comments (that may or may not explain the issue):
(1) Group effects can appear significant even if the amplitude for each subject is very low. This may make it hard to see the FIR effects in some subjects.
(2) FIR needs all trials to begin at the same time relative to the beginning of the TR. If you have some trials that begin at the mid-point of the TR and some the begin at the beginning, then they need to be coded as separate conditions.
(3) FIR needs all trials to be the same duration. Differing durations will hurt the FIR model as the FIR model cannot take duration as an input like the canonical HRF analysis does.
(4) 7 time bins seems quite short. I'm not sure how MarsBar compute time bins or what it recommends (that's a question to ask on the MarsBar listserv). For FIR to work well, you want to model out to 20-30 second for each trial. Thus, the window length should be set to 20-30 seconds, and the order should be set to the number of TRs the occur in that time period. Rather than plotting time bins, you should plot time after stimulus onsets. Generally, if the FIR is setup correctly, it should give you results where you can pick out the response - if the HRF amplitude is great enough.

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 Mon, May 5, 2014 at 1:17 PM, Colin Hawco <[log in to unmask]> wrote:
I have also observed that my attempts to derive estimated time courses often don't look much like HRFs when you average within a group. I think at least part of this is because of the rather high variability in HRFs. Some are large, some small, some short, some long, and some people peak very early or late (I can say based on some of my past work we saw peaks ranging from at least 3 to 9 seconds). 

I think these sources of variability contribute to difficulties in extracting a nice clean time-course. I had had some success in the past extracting person-specific HRFs (using a fourier basis set, and even then about 405 of the time the we rejected the HRFs as to noisy). But averaging together always makes a mess for me. Having lots and lots of participants might help. 

There may be other issues related to FIR which others can comment on, though I have never tried this. 

Colin


On 5 May 2014 04:47, Andy Yeung <[log in to unmask]> wrote:
To supplement information, as a result of my steps, I extracted timecourse from first level of every individual, then transfer to Excel and plot average curves.


On Mon, May 5, 2014 at 4:44 PM, Andy Yeung <[log in to unmask]> wrote:
Dear all,

I used canonical HRF model (the default way) to specify 1st level for every subject, then in 2nd level ROI analysis I used 1-sample t-test to test for significance for each of my conditions (condition1 & condition2).
Results showed I got significant peak voxels (pFWE <.05) for condition2 (at ROI 1 & ROI 2) and for condition1 (at ROI 3).

To further explore, I used FIR model to divide my timecourse into 14 bins, 7 bins for each condition (condition1 & condition2). I extract beta value with Marsbar by the following method:

Attached are my resulted plots. Shapes are not similar to classical HRF curves, could anyone explain why this is so? And if I want to plot a graph more conforming to the shape of a classical HRF curve, what can I do? Thanks.

Best,
Andy