Stephen,

I don't think the two tests are equivalent (although I could be wrong).

1) Subtract pre and post scans using fslmaths and then use randomise to do unpaired t-test (Donald's suggestion)

or

This tests is the difference between scans differs by group - at least for an unpaired/two-sample t-test. It provides the statistical test to conclude the interaction exists. A one sample t-test of each group would be the same as below.

2) Perform 2 separate paired t-tests (one t-test for each intervention group) using randomise, to see if one is significant and the other not (Steve's suggestion).

This tests whether or not the pre- and post-tests are different within each group. If you find an effect in one group, but not the other group, you can not conclude the pre-post difference is different between groups. One group could be slightly above threshold and the other group slightly below the threshold such that there is an apparent group difference when one might not actually exist. 

This issue is well-described in: 

Nieuwenhuis, S., Forstmann, B. U., & Wagenmakers, E.-J. (2011). Erroneous analyses of interactions in neuroscience: a problem of significance. Nature Publishing Group, 14(9), 1105–1109. doi:10.1038/nn.2886

Best Regards, Donald McLaren

On Wed, Sep 17, 2014 at 4:23 AM, Stephen Smith <[log in to unmask]> wrote:

Hi

On 17 Sep 2014, at 09:19, Li, Lucia M <[log in to unmask]> wrote:

Dear Steve and Donald,

Thank you both very much for your suggestions; we really appreciate your taking the time to help us! 

So, have we understood correctly that there are 2 possible ways to do this:

1) Subtract pre and post scans using fslmaths and then use randomise to do unpaired t-test (Donald's suggestion)

or

2) Perform 2 separate paired t-tests (one t-test for each intervention group) using randomise, to see if one is significant and the other not (Steve's suggestion).

We will try them both!

they are equivalent - so I would not bother doing both.

Please may we check a couple of basic things?

1) When subtracting pre-post scans in fslmaths, are these scans the filtered_func_data (from a melodic ICA) or raw BET'd fMRI data?

neither - the spatial maps output by dual-regression (ie the same thing as you normally input to randomise) - see the FSL wiki for details.

2) The input files into the paired t-test randomise command is a 4D file of the filtered_func_data (from melodic ICA)?

3) In the randomise command, can the 'mask' argument refer to any region of interest e.g. DMN/hippocampus/a melodic_ic 4D file?

yes

Cheers

Many thanks again to you both for your help and suggestions!

Kind regards,

Lucia

---

From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of MCLAREN, Donald [[log in to unmask]]
Sent: Tuesday, September 16, 2014 3:03 PM
To: [log in to unmask]
Subject: Re: [FSL] resting state analysis: 2 groups, 2 time points

If you subtract the pre and post intervention scans, then you can use a two-sample test with randomise to compare group differences.

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.

On Tue, Sep 16, 2014 at 4:54 AM, Lucia M Li <[log in to unmask]> wrote:

Dear FSL users & experts,

We are hoping to seek advice on conducting the following analysis:

* 2 groups - intervention vs placebo
* both groups scanned pre and post intervention

We would like to see if there is a difference in resting state connectivity (DMN and hippocampus) after intervention, as compared to placebo.

So far, we have constructed a GLM as per the "2-way mixed-effect ANOVA" on the GLM page (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29).

We are a little confused because it says that "the FEAT model cannot be used with randomise" but we have no task data, only resting state. Therefore, we have not performed a first-level FEAT analysis. So, perhaps we should just use this GLM in the dual_regression command line to perform this resting state analysis? We would be using the filtered_func_data from a melodic ICA as the inputs.
e.g.: dual_regression <DMN_mask> 1 <GLM.mat> <GLM.con> 5000 <output> <subj 1 pre filtered_func_data> <subj 1 post filtered_func_data> ... etc.

Thank you in advance for your help.

Kind regards,
Lucia Li
Visiting research student
IBIC, National Centre for Neurology & Psychiatry
Tokyo

---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve

---------------------------------------------------------------------------

Stop the cultural destruction of Tibet