no,
this determines the ROI of the analysis (which is a volume, that can include the whole brain OR a particular cortex OR ...).
in case of no mask as input, randomise will determine the mask by analyzing (simple thresholding) the concatenated 4D input volume.
cheers

On Sun, Feb 1, 2009 at 10:21 PM, Alexander J. Shackman <[log in to unmask]> wrote:
one more related question,

is the -m <mask> file flag for randomise inclusionary (1=keep) or exclusionary (1=drop)?

thanks, alex


On Sun, Feb 1, 2009 at 2:57 PM, Alexander J. Shackman <[log in to unmask]> wrote:
cool - thanks a million, reza.

(it might be worth updating the online documentation to specify that explicitly)

cheers, alex


On Sun, Feb 1, 2009 at 2:54 PM, Reza Salimi <[log in to unmask]> wrote:
yup,
you got it :)
cheers


On Sun, Feb 1, 2009 at 8:39 PM, Alexander J. Shackman <[log in to unmask]> wrote:
hi reza --

i think maybe i have figured out the problem. as noted in my first email, i ran the command as:


randomise -i MERGE.n1peak_NoRimYsLog_
TSDiff.img -o n1peak_NoRimYsLog_TSDiff_p01 -n 5000 -1 -T -c 0.01 -C 0.01

so that what is termed <thresh> in the documentation (http://www.fmrib.ox.ac.uk/fsl/randomise/index.html) is set to .01, ie, to a p-value.

your helpful emails make me think that this is incorrect, and that <thresh> should be specified as a t-value, such as 2.744.

is this the problem? should <thresh> be specified as a t-value rather than an uncorrected p-value?

(this would potentially seem to explain why the clusters are so much larger than i expected in the output maps)

thanks much for your patience and assistance,
alex


On Sun, Feb 1, 2009 at 2:16 PM, Reza Salimi <[log in to unmask]> wrote:
for more info you can read: http://www.fmrib.ox.ac.uk/fsl/randomise/index.html
(which also mentions why you have a very extended signal, because of high sensitivity of cluster-mass inference)

in randomise you have num_perm repeats of the same process, where one of these is your real study design, and others (num_perm-1) are permuted designs.
For each of these, you have one tstat image.
Then, you compare the real design's tstat image (also saved in the result folder) with the pool of permuted-design's results.

in the cluster-mass inference, real design's tstat is thresholded to result in some clusters a subset of which will be significant,
AND THIS THRESHOLD ID NOT A P-VALUE, it is a t-value (i.e., 2.5)
that is what I said.

Say, you formed the clusters by using 2.3,
then, take tstat1, use the same 2.3 and threshold it,
your suprathreshold *clusterm_corrp* is a subset of this.

cheers


Now, in the result folder, your _corrp_ maps are those corrected for multiple comparison, SO, you dont need to correct it yourself.


On Sun, Feb 1, 2009 at 7:30 PM, Alexander J. Shackman <[log in to unmask]> wrote:
Hi Reza (et al),

Thanks for the rapid response. My residual comments/questions follow,


"the other issue here is that tmap stired in the result folder is different from a tmap or a zmap you have in a parametric inference, where you can convert to p-value and do an inference.
this tmap is the result of the model fitting of the REAL (as opposed to permuted) design, which should be (vox-wise, cluster-wise or cluster-mass-wise) compared to a null dist to give a p-value map.

So, your resulting map is a subset of tmap>(cluster-forming-threshold) NOT tmap>(un)corrected-pvalue"

I understand your comment about the clustermass_corrected-p map NOT necessarily forming a subset of the thresholded t-map, given that both extent and height are part of the equation. I should have realized that -- sorry!

But, I do *not* understand your comment about the t-map. The suprathreshold cluster mass corrp map is not a subset of the corrected t-map. Likewise, the suprathreshold t-map is not a subset of the suprathreshold cluster mass corrp map. They seen to be only marginally overlapping, mainly because the suprathreshold cluster mass corrp map contains most of the brain (whereas the suprathreshold t-map contains only a few moderate sized clusters). 

Relatedly, the values in the t map ("n1cccc_NoRimYsLog_TSDiff_p01_tstat1.nii.gz") created by randomise span a wide range of values -- from -3.71 to +5.04 with every voxel taking a nonzero value. This indicates that it is not thresholded, nor a subset of the t-map formed from fitting the glm (cond1 vs cond2). 

So my question is what map or combination of maps should be used to apply FWE-corrected thresholding?

Thanks for your patience,
Alex Shackman

--
Alexander J. Shackman, Ph.D.
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: +1 (608) 358-5025
Fax: +1 (608) 265-2875
Email: [log in to unmask]
http://psyphz.psych.wisc.edu/~shackman




--
G. Salimi-Khorshidi,
D.Phil. Student, Dept. of Clinical Neurology, University of Oxford.
[log in to unmask]    http://www.fmrib.ox.ac.uk/~reza
FMRIB Centre, John Radcliffe Hospital,
Headington, Oxford,  OX3 9DU
Tel: +44 (0) 1865 222466  Fax: +44 (0)1865 222717




--
Alexander J. Shackman, Ph.D.
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: +1 (608) 358-5025
Fax: +1 (608) 265-2875
Email: [log in to unmask]
http://psyphz.psych.wisc.edu/~shackman



--
G. Salimi-Khorshidi,
D.Phil. Student, Dept. of Clinical Neurology, University of Oxford.
[log in to unmask]    http://www.fmrib.ox.ac.uk/~reza
FMRIB Centre, John Radcliffe Hospital,
Headington, Oxford,  OX3 9DU
Tel: +44 (0) 1865 222466  Fax: +44 (0)1865 222717




--
Alexander J. Shackman, Ph.D.
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: +1 (608) 358-5025
Fax: +1 (608) 265-2875
Email: [log in to unmask]
http://psyphz.psych.wisc.edu/~shackman
Calendar {still under construction}: http://www.google.com/calendar/embed?src=ajshackman%40gmail.com



--
Alexander J. Shackman, Ph.D.
Laboratory for Affective Neuroscience
Waisman Laboratory for Brain Imaging & Behavior
University of Wisconsin-Madison
1202 West Johnson Street
Madison, Wisconsin 53706

Telephone: +1 (608) 358-5025
Fax: +1 (608) 265-2875
Email: [log in to unmask]
http://psyphz.psych.wisc.edu/~shackman
Calendar {still under construction}: http://www.google.com/calendar/embed?src=ajshackman%40gmail.com



--
G. Salimi-Khorshidi,
D.Phil. Student, Dept. of Clinical Neurology, University of Oxford.
[log in to unmask]    http://www.fmrib.ox.ac.uk/~reza
FMRIB Centre, John Radcliffe Hospital,
Headington, Oxford,  OX3 9DU
Tel: +44 (0) 1865 222466  Fax: +44 (0)1865 222717