 |
 |
Maybe you can use %BOLD instead as a measure of effect size.
Matt.
From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Thomas Nichols <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Thursday, February 14, 2019 at 5:04 AM
To: "[log in to unmask]" <[log in to unmask]>
Subject: Re: [FSL] Converting t-statistics to Cohen's d?
Hi Nicole,
Getting a Cohen's d for an arbitrary contrast out of a GLM is basically impossible. That's because Cohen's d is only precisely defined for one-sample and two-sample comparisons.
I can offer three options:
1. One-sample t-test. If your design is nothing more than a one-sample t-test, then you can use this relationship to compute Cohen's d from the t statistic image *_tstat*
d = t / sqrt(N)
2. Two-sample t-test. If your design is just a two-sample t-test of groups of size N1 and N2, and your contrast tests for the difference, then you can use
d = t * sqrt( 1/N1 + 1/N2 )3. For any other contrast, you need to work with R^2 or Cohen's f^2. These relationships will get you there. Make an F test if you don't already have it, as F = t^2, and then
f2 = F * v1/v2R2 = f2 / ( f2 + 1 )
where v1 is the numerator DF and v2 is the denominator DF of your F-test; for a squared t-test, this is v1=1, v2=N-p, where p is the total number of regressors in your model.
f2 is the noncentrality parameter used for power analyses, while R2 has the more natural "variance explained" interpretation.
-Tom
Hi - Cohen's D is the effect divided by the standard deviation of the errors, whereas varcope is similar but is the stddev is divided by the degrees of freedom, so that it reflects the uncertainty of the cope. To get back to D you would need to divide cope/varcope by the temporal degrees of freedom. The *exact* calculations here are slightly complicated by adjustmet for temporal smoothness in the prewhitening step.
Cheers.
On 13 Feb 2019, at 21:09, Nicole Petersen <[log in to unmask]> wrote:
Hi Matthew (and others),
Thanks for this recommendation. I have done this (finally) and then divided the cope images by sqrt(varcope) to produce what I believe are effect size maps in Cohen's D. However, some of the Cohen's D values seem improbably high -- higher than 4. (These maps are being used to visualize differences that were NOT statistically significant, so I was expecting small Cohen's D values.) Questions:
1. Is my math wrong, and there is another way to get Cohen's D from these images?2. If my math is right, does this also seem improbably high to you?
Thanks in advance for any advice.
Best,Nicole
Hello,
You can use the --glm_output option to output this glm info directly from randomise.
Kind Regards
Matthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
> On 26 Nov 2018, at 20:38, Nicole Petersen <[log in to unmask]> wrote:
>
> Hi FSL experts!
>
> I have been tasked with creating effect size maps in Cohen's d out of randomise outputs. I understand that this could be done by using film_gls --noest to produce PE and COPE images, which can then be converted to Cohen's d through fslmaths. I'd like to try converting the t-statistic images produced by randomise directly (since those are the analyses we're reporting on). Do you all have recommendations as to how to compute these maps?
>
>
> Thanks for any help,
> Nicole Petersen
>
> ########################################################################
>
> 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
--
Nicole Petersen, Ph.D
Postdoctoral Fellow
London Lab
Semel Institute for Neuroscience and Human Behavior
Department of Psychiatry and Biobehavioral Sciences
University of California, Los Angeles
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Head of Analysis, WIN (FMRIB) Oxford
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 610470
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------Stop the cultural destruction of Tibet
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
--
__________________________________________________________
Thomas Nichols, PhDProfessor of Neuroimaging StatisticsNuffield Department of Population Health | University of Oxford
Big Data Institute | Li Ka Shing Centre for Health Information and Discovery
Old Road Campus | Headington | Oxford | OX3 7LF | United Kingdom
T: +44 1865 743590 | E: [log in to unmask]
W: http://nisox.org | http://www.bdi.ox.ac.uk
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
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