[Homer-users] Using the GLM_HRF_Drift_SS function in HomER

[Meryem Ayse Yucel]

Hi Madeleine,

The reason you are getting very smooth HRFs is that you are using the
default option for GLM_HRF_Drift_SS function which is fitting a modified
gamma function to the data (idxBasis = 2). The shape is fixed to a gamma
function and only the amplitude is estimated by GLM. Instead you can use
the first option (idxBasis =1) which uses a sequence of gaussian functions
and is therefore more flexible. Parameters for this option can be [1 1]
(width and temporal spacing of the gaussian respectively.
Also note that you can hover over the parameter names in options menu to
get more information for each.

Hope this helps.

Meryem



> Hi,
>
> We have a few questions about using the GLM_HRF_Drift_SS function in
> HomER. We've had a quick look through previous forum posts and have not
> found a clear answer.
>
>   1.  From our understanding, if we run a basic processing stream that
> includes the GLM_HRF_Drift_SS function without a Block_Average function
> afterwards, what we see when we look at the plot after clicking on 'show
> Run HRF' is the predicted response, not our actual data (and this is why
> it looks so 'perfect' and smooth, and why it changes according to the
> parameters that we choose). On the other hand, if we add a Block_Average
> function after the GLM_HRF_Drift_SS function in our processing stream,
> what we see when we look at the plot after clicking on 'show Run HRF' is
> the actual response (which is why it looks less smooth and more
> natural). Is this correct? If not, could someone please explain what the
> GLM_HRF_Drift_SS function is doing?
>   2.  If this is correct, what options do we have in terms of using the
> model (i.e. the predicted data)? Is it possible to compare it with the
> actual data and to obtain a goodness of fit value? Or would we need to
> export the model and data and do this ourselves?
>   3.  Finally, are there any rules of thumb for choosing the parameters
> for the model? For instance, idxBasis options 2-4 require three
> parameters: delay, width, and stimulus duration - are there any
> recommendations for the first two? We could base the decision on
> previous data, but is it recommended to try a few different parameter
> values to see which fits best with the actual data?
>
> Many thanks in advance for your help!!!!
>
>
> Madeleine
>
>
> -----
>
> Madeleine Verriotis, PhD
> Chair, Neuroscience Careers Network
> Postdoctoral Research Associate
>
> UCL Dept. of Neuroscience, Physiology and Pharmacology
> Tel: 020 7679 3533. (internal: 33533)
> Email: madeleine.verriotis at ucl.ac.uk
> _______________________________________________
> Homer-users mailing list
> Homer-users at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/homer-users