[Neuroimaging] GC analysis with nitime

[Ariel Rokem]

Hi Mike,

Thanks for your email. Answers inline below:

On Tue, Jul 9, 2019 at 3:12 PM Mike V <mikeltv95 at gmail.com> wrote:

> Thank you Bennet, I'll wait for the next release then.
> Best regards,
> Mike
>
> On Sat, 6 Jul 2019 at 14:18, Bennet Fauber <bennet at umich.edu> wrote:
>
>> Mike,
>>
>> I reported the blank plots in an Issue at GitHub.  I think there may
>> be more that is needed before they make a new release and refresh the
>> tutorial page.
>>
>>     https://github.com/nipy/nitime/pull/176
>>     https://github.com/nipy/nitime/issues/173
>>
>> Sorry, can't help with the rest.
>>
>> On Fri, Jul 5, 2019 at 7:37 PM Mike V <mikeltv95 at gmail.com> wrote:
>> >
>> > Hello,
>> >
>> > I'm following the Granger Causality tutorial in the nitime's webpage
>> but I have few questions:
>> >
>> > 1- In the tutorial the values f_ub = 0.15 and f_lb = 0.02 are used for
>> the bounds on the frequency band of interest. Are these values recommended
>> for all resting state data-sets or do they depend on the particular
>> acquisition parameters?
>>
>
These are pretty reasonable for most fMRI data, with anything above 0.15 Hz
probably unrelated to the hemodynamic response and slow oscillations below
0.02 probably due to signal drift (but we also have a paper coming out soon
that there are some interesting things going out in that infraslow range).
But it really depends on your data and on your questions.


> >
>> > 2- I do not understand the causality plots (fig03 and fig04), they are
>> both blank. What does it mean? I was expecting some variation like in the
>> coherence and correlation matrices...
>> >
>>
>
Yes. As Bennet pointed out, this is a bug in the recent release. I really
hope that we can fix that soon, but after a quick look, I am not quite sure
what's going on, and have only limited time to devote to this in the next
few weeks.


> > 3- the tutorial shows the analysis of one single subject. How can I
>> compute group statistics? Can I simply extract the values stored in g1 and
>> then run a one sample t-test across all subjects' g1 scores?
>> >
>>
>
That seems like a reasonable approach.


> > 4- how can I get GrangerAnalyzer to display significance?
>> >
>>
>
It's not designed to do so.


> > 5- I noticed that the tutorial data has been motion corrected only. Is
>> it not necessary to correct for other sources of noise like physiological
>> noise and linear drift before computing GC?
>> >
>>
>
The filtering should help with that (see above).

Hope that helps -- sorry about the slowness to respond and fix things,

Ariel


> > Many thanks in advance!
>> >
>> > Best regards,
>> > Mike
>> > _______________________________________________
>> > Neuroimaging mailing list
>> > Neuroimaging at python.org
>> > https://mail.python.org/mailman/listinfo/neuroimaging
>> _______________________________________________
>> Neuroimaging mailing list
>> Neuroimaging at python.org
>> https://mail.python.org/mailman/listinfo/neuroimaging
>>
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20190709/e86e9459/attachment.html>