Print

Print
Hi Ludovico,

For (1): the temporal filter is something that appears in the Melodic GUI but it's used only in the preprocessing. If you use GM maps, you'd use the "melodic" command line, and there is no temporal filtering there (that is, for the ICA algorithm itself).

For (2): the 0.5 isn't like a "p-value" in the usual sense. It's more a trade-off between the probabilities of false positives and false negatives. You can keep or change, though I think it's more common to give equal weight, so, 0.5 then is the usual.

All the best,

Anderson


On 23 February 2017 at 19:03, Ludovico Coletta <[log in to unmask]> wrote:

Hi Shane,


thank you for the answer. 


Regarding the threshold: "A threshold level of 0.5 in the case of alternative hypothesis testing means that a voxel 'survives' thresholding as soon as the probability of being in the 'active' class (as modelled by the Gamma densities) exceeds the probability of being in the 'background' noise class. This threshold level assumes that you are placing an equal loss on false-positives and false-negatives" (from the melodic wiki). 

In my opinion a threshold of 0.5 is too low, but maybe I'm only paranoic


Best

Ludovico



Da: FSL - FMRIB's Software Library <[log in to unmask]> per conto di Shane Schofield <00000fa5b2845489-dmarc-[log in to unmask]>
Inviato: giovedì 23 febbraio 2017 17.41
A: [log in to unmask]
Oggetto: Re: [FSL] Grey matter networks in FSL - Follow up
 
Hi Ludovico,

Thanks for following up on this.

1. I think it is 1, too. Look forward to Anderson’s take.
2. Can you explain what is this threshold about?

Cheers,
Shane

-- 
Shane Schofield

On 23 February 2017 at 14:49:48, Ludovico Coletta ([log in to unmask]) wrote:

Dear FSL experts,

I followed with great interest the discussion about the usage of MELODIC and the possibility to analyse the integrity of GM networks. 
I have two (perhaps trivial) questions about the settings:

1) What is the right cut off for the high pass filter when the inputs are structural GM maps? I think it should be 1.0 s, since each "fake" time point is independent

2) Any advise on the threshold for the IC maps? Previous studies used 0.5, which I think is too low. What about 0.8?

Thank you for your time
Ludovico