Difference in FWE and FDR thresholds between xjView and SPM8

[ckennel]

I'm comparing xjView's volume report with SPM8's whole brain statistics report and noticed a discrepancy that I haven't found an explanation for. 

I'm analyzing the exact same spmT file with the same p thresholds (p=0.001), cluster size (5 voxels) and search volume (whole brain), and in the legend at the bottom of the statistics table all values are the same except the FWEp, FDRp, FWEc, and FDRc.  Here is what each program gives:

xjView: FWEp: 4.707, FDRp: 4.294, FWEc: 338, FDRc: 338
SPM8: FWEp: 8.792, FDRp: 8.512, FWEc: 300, FDRc: 20

Since the rest of the content in the legend and table above is the same between both programs, I'd like to know why the corrected T thresholds vary as it affects how I interpret the results. 

Thanks in advance,
Chris Kennel

[Xu Cui]

Chris,

The 'volume' feature is mainly for users to conveniently find the list of clusters and their peak locations; the statistics might not be accurate as it depends on other information (e.g. SPM.mat). I myself usually don't look at the statistics.

The general rule is, if you find a discrepancy, you should use the result of spm.

Xu

--
You received this message because you are subscribed to the Google Groups "xjView discuss group" group.
To view this discussion on the web visit https://groups.google.com/d/msg/xjview-discuss-group/-/3rDn_1a1yW0J.
To post to this group, send email to xjview-dis...@googlegroups.com.
To unsubscribe from this group, send email to xjview-discuss-g...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/xjview-discuss-group?hl=en.

--

----------------------------------------------------------------
Xu Cui, Ph.D. 崔旭 博士
Department of Psychiatry and Behavioral Sciences
Stanford University
Blog: http://www.alivelearn.net

Profile: http://www.peaya.com/p/?cuixu
----------------------------------------------------------------

[ckennel]

Thank you very much for taking the time to answer my question, and thank you also for making xjView. 

I have one other question: is it possible to set the middle of the colormap to always display a particular color for a particular T value or intensity without having to manually go through the tedious process of editing the colormap in the GUI?  I would like to have a consistent color (green) centered on 0 regardless what my range of T values is.  That way positive T values always display with warm colors and negative T values with cool colors.  Some of my data has T values distributed evenly around 0 while many other files have a larger range in the positive direction.  There is too much variation in range from one file to another to allow me to save a few colormaps and apply them to the figures.  Ideally I'd like to be able to set the maximum, middle, and minimum values of the colorbar to keep it constant between images.

All the best,
Chris

[Xu Cui]

Chris,

The current version of xjview unfortunately doesn't do that. But I agree it would be a very handy feature.

However, here is a trick you can do:

In xjview.m, around line 9823, find:

mx = str2num(TMAX_);

Then add the following line right below it:

mn = -str2num(TMAX_);

Then you simply set colorbar max to a certain number, the section view will have a consistent symmetric color.

Xu

--
You received this message because you are subscribed to the Google Groups "xjView discuss group" group.

To view this discussion on the web visit https://groups.google.com/d/msg/xjview-discuss-group/-/aqYUKYN1VrUJ.

To post to this group, send email to xjview-dis...@googlegroups.com.
To unsubscribe from this group, send email to xjview-discuss-g...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/xjview-discuss-group?hl=en.

[ckennel]

Xu,

Thank you so much again.  That works perfectly for me, and now all my images have a standard look.

Chris

[ckennel]

I think I figured out the discrepancy between the FWE and FDR thresholds in SPM8 and xjView.  It appears to me that xjView reports a threshold z value while SPM8 reports a threshold t value.  Looking at the p FWE-corr and q FDR-corr columns in the volume reports created by both programs seems to back this up; though I'm sure it's written down in a manual somewhere (I just haven't come across it yet).

All the best,
Chris

[malak abu shakra]

Dear Dr. Cui, 

I have a question related to xjview interface: for some reason, I'm not managing anymore to see x y z co-ordinates for a given con images nor am I able to enter the co-ordinates I want in the boxes. It is as if the xjview interface stops at the pValue specification point. I don't think it's my computer since I've tried this on 3 different ones, and also tried to re-download xjview, all with no success. 

Your advice would be greatly appreciated !

Malak

Malak Abu Shakra    Doctoral Student
Clinical Psychology
Psychology Dept.   Robert O. Pihl Lab
McGill University
1205 Dr. Penfield Avenue
Stewart Biology Building - Room W8/37
Montreal, Quebec, H3A 1B1
Canada

"The most erroneous stories are those we think we know best - and therefore never scrutinize or question". 

Stephen Jay Gould - 

[Xu Cui]

Hi, Malak,

Sorry for the long delay. For some reason google group has held your post for a while.

I understand the problem you described but unfortunately I have no idea how it happened and how to fix it. If anybody in this group knows a solution please help.

Xu