Thanks Anderson. I see that Woo argues for 3.1. Paradigms with complex cognitive tasks though, don't really come up to that significance level. Does the random field theory hold at 2.3? 

Best,

Julian

On Wed, 3 Apr 2019, 17.25 Anderson M. Winkler, <[log in to unmask]> wrote:

 Hi Julian,

The default is now 3.1. The paper by Woo et al (Neuroimage, 2014) made a good case for that. I don't think the threshold level was on its own right a cause for increased amount of false positives in any tool that was assessed, although too low thresholds do indeed create problems (the random field theory stops working, for example).

All the best,

Anderson

On Wed, 3 Apr 2019 at 05:31, Julian Macoveanu <[log in to unmask]> wrote:

Dear FSL experts,

In the light of the Eklund papers, I was wondering what is the recommended cluster forming threshold for group level estimation using the FLAME1(+2).

I think the default is z=2.3 and that 80% of FSL papers use this according to the latest survey doi.org/10.3389/fnhum.2018.00016, but Eklund and others argues for 3.1. However, they use an older FSL version 5.07 and 5.09 and z=3.1 is more a general recommendation for least-square based estimations methods, and not sure to what degree it should applied to FLAME which is a much fancier mixed-effect model estimation.

It is easy to recommend a more stringent CFT, but 3.1. seems rather conservative using FLAME and looks to me to result in high false negative rates.

The question is, what CFT in the current FSL version should we choose to get a rate of 5% false positive rate that we want to accept at corrected p<0.05.

All the best,

Julian

---

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

---

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

---

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1