Hi Doug,
Sorry to follow this topic with a delay but we tried to model BOLD
signal undershoot with a gamma function and we failed. The problem is,
gamma function generates positive values and we can NOT use inverse gamma
function (or otherfunction similar to that) as a part of mkanalysis-sess
What do you mean the gamma function generates positive values? It will
create a regressor with a positive-going HRF. However, the regression
coefficient can be negative which would invert the shape. The reg coef
is fit to the data, so the data will tell you whether there is an
undershoot or
You don't generate a negative coefficient. The sign and magnitude of the
coefficient is determined by the fitting procedure. It may be the case
that a positive coef fits the data better than a negative. You should
not change the gamma parameters.
doug
On 04/24/2013 04:06 PM, SHAHIN NASR
Sorry that I keep on asking questions but I want to know if using a model
can significantly increase the chance of getting a significant response.
As far as I see for the positive peak, when I use a model (e.g. gamma
model or spmhrf), I see a more significant response compared to when I use
Yes, assuming a shape will give you much more power, at least for the
individual subject.
doug
On 04/11/2013 02:55 AM, SHAHIN NASR wrote:
Sorry that I keep on asking questions but I want to know if using a
model can significantly increase the chance of getting a significant
response.
That's the problem Doug. I have already generated the FIR model but how can
I show the sig MAP for one particulat time point. What are the other
ways?
On Wed, Apr 10, 2013 at 4:05 PM, Douglas N Greve
gr...@nmr.mgh.harvard.eduwrote:
Hi Shahin, there are several ways that you could do it. The
yes, that is what you want. The only other way that comes to mind is to
model each event type as two event types one delayed relative to the
other. Then model the responseto each as a gamma. If the delay is right,
then the first gamma should model the main positive response and the
second
