Dear Donald,
I was really looking for a detailed explanations as what you did.
Can you please comment from your view in the event design that I mentioned earlier?
If my aim is to get both : the highest possible detection, which I think better using a block design, and want at the same time to see the effects of linearity and non linearity, will my event design that I will do it soon is the best to be used?
To save your time this is what I want to do in brief:
1) Jittering the events and on average the ISI will be 4s.
2) The length of a run will be ~ 8 min
3) I will use TR of 2.5, TE 40
4) the duration of an event will be ~1.7s ( if I made it less than this what the difference is that I may find)
5) the number of events will be around 100 and the rest among them will be ~80
Your detailed comments will be really appreciate and thanks again for your last message.. I was looking for an answer like this for around a month:-)
AS
On 16 May 2013, at 22:39, "MCLAREN, Donald" <[log in to unmask]> wrote:
> Chris and AS,
>
> I'd agree that an event-related design would be much better suited for
> the question of linearity of squeeze strength. However, I think it
> could be modelled with events as I will explain below.
>
> If you have a block of 30s, you'd model it as 1 event of 30s. Now I
> would get the same results if I modelled it as 3 events each lasting
> 10s with no gap. On the extreme, I could model is as 30 events each
> with a duration of 1s. All 3 models would produce the same regressor
> and thus the same estimate. Although you've modelled it as short
> events, you still have the block regressor. Now, we can add a
> parametric modulator that might explain some of the variance in
> squeeze strength. The two potential issues are: (1) the variance of
> squeeze strength is below the variability of the response during the
> block; (2) the parametric modulator is mean-centered per run; and (3)
> there might be additional time effects (e.g. habituation in each
> block). Thus, if each run doesn't contain enough of the high and low
> squeeze strengths, then you might not see much of an effect.
>
> Modeling random events is not the way to go as you are telling the
> program that there were no squeezes during the skipped events. This
> does not provide any basis for what a true event-related study would
> show.
>
> 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 Thu, May 16, 2013 at 2:36 PM, Chris Watson
> <[log in to unmask]> wrote:
>> I wasn't surprised because you are trying to create an event-related model
>> with trials that are too close in time. What you plan on doing in the future
>> sounds fine; just be sure to include a random jitter (e.g. sometimes the ISI
>> is 3s, sometimes it is 5s, etc.)
>>
>> On 05/15/2013 04:51 PM, fMRI wrote:
>>
>> Dear Chris,
>>
>> Can you please tell me why you were not surprise as this is what I want to
>> understand?
>>
>> Can you please recommend a event related design for me to get what I want in
>> the most efficient way.
>>
>> I will share with you what I am thinking to do next time and I will
>> appreciate your comments:
>>
>>
>> I will do an 8 min experiment with a TR of 2.5 s. I will make it event
>> related such that the average ISI will be ~ 4 s. I will try to have ~ 100
>> events. Do you expect that I get something better using this design? Any
>> comments will be appreciate it ?
>>
>> Thanks
>>
>> AS
>>
>> On 15 May 2013, at 21:18, Chris Watson
>> <[log in to unmask]> wrote:
>>
>> 1) That doesn't surprise me that your results were not as expected.
>> 2) I still don't think this should be modeled as an event-related design.
>> The stimuli are too close together. I don't know how you could justify
>> selecting a random subset of stimuli to include in your design and then run
>> the analysis.
>>
>>
>> On 05/15/2013 04:06 PM, fMRI wrote:
>>
>> The design is like this:
>>
>> 1) 48 blocks, 24 task and 24 rest
>> 2) the subject was squeezing in the active task. Each block has a different
>> target so that the strength in each block is different.
>> 3) the isi was 0.6 second between each squeeze.
>> 4) I had around 320 volumes and the TR was 3.1
>> 5) each block lats for 20 seconds
>>
>> I design it in two ways:
>>
>> 1) the condition was the onset time of each squeeze and the duration was
>> zero so as a delta function. Then I added the grip strength values for each
>> onset in the parametric modulation.
>>
>> The result from this one was not as I expected as the linearity was not high
>>
>> 2) I thought that if I want to see the parametric effects , I should use an
>> event design. So before doing this I tried to select randomize onset with
>> its grip strength, such that the average isi is 4 seconds. I used the same
>> number volumes without changes. I got here very nice results.
>>
>> Any comment will be appreciate it,
>>
>>
>> Thanks
>>
>>
>> On 15 May 2013, at 19:03, Chris Watson
>> <[log in to unmask]> wrote:
>>
>> I think we'll need to know more about your design, e.g. stimulus duration,
>> inter-stimulus interval, and so forth.
>>
>> On 05/15/2013 01:01 PM, fMRI wrote:
>>
>> Hi all,
>>
>> I want to understand basic thing about spm. What is really more important to
>> spm the active or task volumes or the specification of a condition.
>>
>> For example, I have a block design and then defined it as event by randomly
>> selecting a number of trials among the active and rest blocks. I get ~ same
>> activations using the exact same volumes but changing the condition from ~
>> 1000 onset to 100 onsets. The reason why I did that was because I did an
>> experiment to see the linear effects of using parameters as a mixed block
>> design. The result was not similar to what other did. Then I found that it
>> is better to design an experiment such this as an event design. Since I did
>> not have a new experiment, I thought I can play with what I have. In terms
>> of the linearity, I got really nice result as what I want. I just want to
>> understand how and why although I used the same volumes. Does spm ignore the
>> other activations that I do not defined? Can you comment please ?
>>
>> Thanks
>>
>> AS
>>
>>
>>
>>
>>
|