hi Martin,
in our recent work [1,2] where we estimate a model similar to the one
you mention
we constrain the HRF to peak at 1 which fixes a potential problem for
second level.
Best,
Alex
[1] http://www.sciencedirect.com/science/article/pii/S1053811914008027
[2] http://arxiv.org/abs/1402.7015
On Sat, Jan 24, 2015 at 4:32 PM, Martin Hebart <[log in to unmask]> wrote:
> Dear Donald, dear all,
>
> Thanks a lot for you reply.
>
> 2015-01-23 23:03 GMT+01:00 MCLAREN, Donald <[log in to unmask]>:
>>
>> How did you come up with your subject-specific HRFs? I don't think you
>> want to rescale them, but I could be wrong. It may depend on how they were
>> generated.
>
>
> In fact, I don't only have subject-specific HRFs, but voxel-specific HRFs. I
> get them from the data by fitting both the beta and the HRF in a constrained
> manner (varying only the onset and dispersion parameters in a small range),
> but assuming that the HRF is the same across all regressors. The dispersion
> parameter changes the width and amplitude of the HRF.
>
>>
>>
>> The interpretation is that 1 unit of neural activity would cause a 1 unit
>> increase in the BOLD signal. Thus, if you rescaled the HRFs, then 1 unit of
>> neural activity would not cause a 1 unit increase in the BOLD signal.
>
>
> Ok, but I'm not sure I fully understand. Maybe my misunderstanding stems
> from the idea that I usually think of a beta as estimating the height of the
> HRF whereas of course all points of the HRF contribute to the size of the
> beta. Maybe the question is whether I'm interested in the amplitude of the
> BOLD effect (percent signal change) or in the beta estimates.
>
> Let's say I wanted to calculate something akin to percent signal change,
> then I would usually multiply the peak of the HRF with the beta divided by
> the mean of the data, which - assuming no difference in mean across subjects
> - should effectively have the same result as rescaling the regressor. Does
> this mean that in case I'm interested in the amplitude it would be better to
> rescale?
>
> I normally don't have to choose between the two, but I'm a little confused
> as to which interpretation is more appropriate when it comes to the
> second-level results. Any help would be highly appreciated.
>
> Cheers,
> Martin
>
>
>
>
>
>
>>
>>
>> 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 Fri, Jan 23, 2015 at 1:01 PM, Martin Hebart <[log in to unmask]>
>> wrote:
>>>
>>> Dear all,
>>>
>>> I'm using subject-specific HRFs (from spm_hrf with slightly different
>>> parameters) at the first level in my analysis. Now I want to look at some
>>> contrasts at the second level.
>>>
>>> However, the height of the HRFs from spm_hrf differ across subjects. This
>>> results from the integral of spm_hrf always summing up to 1, i.e. a wider
>>> HRF gives a lower peak amplitude. I believe this makes the betas across
>>> subjects not comparable.
>>>
>>> Is it ok to scale the HRFs in advance that they are all equal in height?
>>> I've been wrapping my head around this issue, but I'm unsure and would be
>>> happy about some confirmation or alternative suggestions.
>>>
>>> Thanks!
>>> Martin
>>>
>>
>
|