Hi Liam -

Are the probabilities changing from trial to trial - I think that is
how TDRLs usually work?  If so, what I think you want to do is have
regressors where the height of the modelled response is modulated
according to the TDRL.  If you use 3-column format timing model, the
height corresponds to the values in the third column, so you can
modulate this.  Depending on your hypothesis, you may want to have
this regressor in addition to a mean response regressor.    A lot of
the decision making studies from FMRIB use this sort of modelling,
without (I believe) a great deal of hacking of FSL.

Eugene


On 19 October 2011 13:23, Liam Nestor <[log in to unmask]> wrote:
> Hi Eugene
>
> I have a monetary incentive delay (MID) task. I want to try and employ a
> temporal difference reinforcement learning (TDRL) model.
>
> The MID has a prepare period, 3 cue conditions and 6 outcome conditions
> (EV=10).
>
> The contrasts generated for each subject, for each run (n=3) from the first
> level analysis are:
> Cope 1 = Loss Cue (EV 1)
> Cope 2 = Win Cue (EV 2)
> Cope 3 = Neutral Cue (EV 3)
> Cope 4 = Lost Outcome (EV 4)
> Cope 5 = Saved Outcome (EV 5)
> Cope 6 = Lost+Saved (EV 1+EV 2)
> Cope 7 = Won Outcome (EV 6)
> Cope 8 = Missed Outcome (EV 7)
> Cope 9 = Won+Missed (EV 6+EV 7)
> Cope 10 = Neutral Lost Outcome (EV 8)
> Cope 11 = Neutral Won Outcome (EV 9)
> Cope 12 = Prepare (EV 10)
>
> The TDRL model concerned involves the Win Cue period X probability (0.44) -
> the "expected value". The "expected value" is calculated as the sum of all
> possible outcomes (4.5) X probability (0.44).
>
> Therefore, I want to run the formula  - Win cue X 0.44 - sum of all possible
> outcomes (4.5) X probability (0.44) to generate a new image file. I have
> tried the following:
>
> fslmaths cope2.nii.gz -mul 0.44 -sub 4.5 -mul 0.44 cope13.nii.gz
>
> It generates cope13.nii.gz, but it won't threshold on this file when I run a
> Fixed Effects level analysis averaging across the 3 runs.
>
> I have no knowledge of being able to use this formula when setting up the
> contrasts in the FSL GLM GUI, which is why I'm attempting to do this in the
> "stats" directory afterwards.
>
> Thanks,
>
> Liam.
>
>
> On 19 October 2011 12:31, Eugene Duff <[log in to unmask]> wrote:
>>
>> Hi Liam -
>>
>> On 19 October 2011 12:14, Liam Nestor <[log in to unmask]> wrote:
>>>
>>> Dear FSL users
>>>
>>> Using fslmaths, I want run a formula, using files that go into the
>>> "stats" directory, which are generated from a first level analysis. The
>>> files that end up in this directory are cope; neff; pe; tstat; varcope and
>>> zstat. The formula is very simple - merely multiplying one of the contrasts
>>> by a number and subtracting another number.
>>>
>>
>> The cope and varcope files are the primary images used in higher level
>> analyses.  Both will be affected by multiplication of the contrasts. You
>> might want to describe want you are trying to do to - I can't really think
>> of a reason why you would want to do what you are saying.
>> Cheers,
>> Eugene
>>
>>
>>
>>>
>>> What I'm not sure about is which contrast files will go into the formula?
>>> The cope; neff; pe; tstat; varcope or zstat files? All of them? I have tried
>>> using the cope, tstat and zstat files in the formula, but when I run a Fixed
>>> Effects analysis (averaging across runs), it will not create thresholded
>>> zstat files for the new files that I have created using fslmaths.
>>>
>>> I'm not able to set this formula up using the FSL GUI when doing the
>>> first level analyses, which is why I am having to resort to this formula
>>> subsequently.
>>>
>>> Thanks,
>>>
>>> Liam.
>>>
>>>
>>> --
>>> Liam J Nestor PhD
>>>
>>
>
>
>
> --
> Liam J Nestor PhD
>
>