Re: Question

[Thomas Nichols]

Dear Ed,

I hope it's OK I'm CCing the SnPM email list.

We did the SnPM analysis using Tools > SnPM > Specify > "2 Groups: Test diff of responses; 2 conditions, 1 scan per condition" to get the correction for our Full Factorial results in SPM8. However, when we went to plug in the SnPM recommended threshold voxel (critical STCT) number and p value into Results with the original SPM.mat file, we saw result regions that were different than those shown in the SnPM output. See pictures 1 and 2 for comparison: "Pic 1_SnPM output_RCL_con01..." and "Pic 2_Results_RCL_con01..." 

Sorry, I'm afraid there was nothing attached.  Also, even without the images, I'm afraid I don't understand exactly what two results you're comparing.

We noticed that the matrices for these two results are different even though we thought we were doing the same thing. However, when we looked into the matrices for the "Flexible Factorial Design" option in SPM described in the link below, it looks the same as the SnPM matrix. Granted, this publication is for SPM version 5, but it looks like the module we used for SnPM is more similar to the flexible factorial analysis than our full factorial analysis. http://www.sbirc.ed.ac.uk/cyril/download/Contrast_Weighting_Glascher_Gitelman_2008.pdf

The one thing that SPM does that SnPM cannot not do is allow for different variances... I wonder if that might be the source of the apparent differences in the design matrix.  In SPM, in flexible factorial, by default different groups have different variances; after the model is estimated, the displayed design matrix reflects the (globally estimated) variance difference between the groups.  This might also produce differences in the statistic maps.

Might you be able to explain this discrepancy between SPM and the SnPM results using the SnPM module ("2 Groups: Test diff of responses; 2 conditions, 1 scan per condition") Specifically, should we be using another method in SnPM to correct our results for the full factorial analysis in SPM or should we be using something other than full factorial in SPM? 

If you can boil your results down to a pair of measurements per subject, this is the right plug in.  And, for what it's worth, it's the same analysis as a two-sample t-test on the differences, if you've pre-computed the intrasubject difference.

-Tom

Best,

Ed

Edward F. Pace-Schott, Ph.D.

Director, Sleep and Anxiety Disorders Laboratory
Assistant Professor of Psychiatry
Harvard Medical School
Massachusetts General Hospital - East
CNY 149 13th Street Room 2605
Charlestown, MA 02129
Phone: 508-523-4288
Fax: 617-726-4078
Email: epace-...@mgh.harvard.edu

The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

[Pace-Schott, Edward F.,Ph.D.]

Hi Tom,

Thanks so much for your quick reply! Sorry, I must not have attached the screen shots. Here they are.

Thanks again!!

Best,

Ed

Edward F. Pace-Schott, Ph.D.

Director, Sleep and Anxiety Disorders Laboratory
Assistant Professor of Psychiatry
Harvard Medical School
Massachusetts General Hospital - East
CNY 149 13th Street Room 2605
Charlestown, MA 02129
Phone: 508-523-4288
Fax: 617-726-4078
Email: epace-...@mgh.harvard.edu

---

From: Thomas Nichols <clne...@nexus.ox.ac.uk>
Sent: Wednesday, July 29, 2020 12:08 PM
To: Pace-Schott, Edward F.,Ph.D. <EPACE-...@mgh.harvard.edu>
Cc: Oliver, Katelyn Irene <KIOL...@mgh.harvard.edu>; Camprodon, Joan A.,M.D.,M.P.H.,Ph.D. <JCAMP...@mgh.harvard.edu>; snpm-support <snpm-s...@googlegroups.com>
Subject: Re: Question

 

        External Email - Use Caution        

[Thomas Nichols]

Ah!  There are two problems.

Firstly, the analysis you're looking at in SPM is an *F* test, testing equality of the two effects in each group, which means you'll reject the null hypothesis if there is a difference in *either* group, which probably is not what you want.

Secondly, there's something strange about your full factorial... usually SPM will create the dummy variables for each subject... I don't see that in your SPM model.  I fear I'm not the SPM Full Factorial expert, but if you scan through that Contrast_Weighting_Glascher_Gitelman_2008.pdf document you'll see design matrices where there are dummy variables for each subject.

Sorry I can't be of more help with SPM, but you *should* be able to get a T test (not F) and a design with subject dummy variables that will be more similar to the SnPM analysis.

-Tom

--
You received this message because you are subscribed to the Google Groups "Statistical Nonparametric Mapping" group.
To unsubscribe from this group and stop receiving emails from it, send an email to snpm-support...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/snpm-support/BL0PR04MB4948E5BB1E0D71D727AE551CDE700%40BL0PR04MB4948.namprd04.prod.outlook.com.

[Pace-Schott, Edward F.,Ph.D.]

Hi Tom,

Thanks so much for your quick reply. We think we're beginning to puzzle this out and we very much appreciate all your advice.

Best,

Ed

                
Edward F. Pace-Schott, Ph.D.

Director, Sleep and Anxiety Disorders Laboratory
Assistant Professor of Psychiatry
Harvard Medical School
Massachusetts General Hospital - East
CNY 149 13th Street Room 2605
Charlestown, MA 02129
Phone: 508-523-4288
Fax: 617-726-4078
Email: epace-...@mgh.harvard.edu

---

From: Thomas Nichols <clne...@nexus.ox.ac.uk>
Sent: Wednesday, July 29, 2020 12:40 PM

To: Pace-Schott, Edward F.,Ph.D. <EPACE-...@mgh.harvard.edu>

Cc: Oliver, Katelyn Irene <KIOL...@mgh.harvard.edu>; Camprodon, Joan A.,M.D.,M.P.H.,Ph.D. <JCAMP...@mgh.harvard.edu>; snpm-support <snpm-s...@googlegroups.com>; Jeehye Seo <amoi...@gmail.com>
Subject: Re: Question

 

[Oliver, Katelyn Irene]

Hi Tom,

We decided to try using the Flexible Factorial design in SPM to see if we could specify a matrix that looked more like the SnPM matrix. I had 3 factors (subject, group, condition-time) and said specify all, using a 32 x 4 matrix to define the row, subject, group, and condition of each of the scan files (8 subjects per group x 2 groups x 2 timepoints = 32 scan files).

I created the SPM.mat file for this design and ran it through results using the SnPM conditions and have attached a screenshot of the comparisons of the SnPM and flexible factorial SPM outputs. (see powerpoint attached to this email).

Although there is still the difference between the f test for the SPM file and the t test for the SnPM analysis, the clusters are looking much more alike and the matrices look almost identical (besides the covariates being in a slightly different location).

Would you say we are on the right track with this? We plan to run some post-hoc t-tests to determine the directionality of the interaction in SPM, but we wanted to get the interaction effect using an f test first. 

I apologize if I've misunderstand what you were trying to explain. Thank you for helping out our group!

Best,

Katelyn and Ed 

---

From: Pace-Schott, Edward F.,Ph.D. <EPACE-...@mgh.harvard.edu>
Sent: Wednesday, July 29, 2020 3:12 PM
To: thomas....@bdi.ox.ac.uk <thomas....@bdi.ox.ac.uk>

[Thomas Nichols]

Yes!  With the exception of using the F-tests in SPM, you've got it.  Examine the T-tests in SPM and you should have exactly the equivalent test.

-Tom

To view this discussion on the web, visit https://groups.google.com/d/msgid/snpm-support/BL0PR04MB4577ACD06162B93C5801152D9F710%40BL0PR04MB4577.namprd04.prod.outlook.com.

[Pace-Schott, Edward F.,Ph.D.]

Hi Tom,

Thank you so much!!

Best,

Ed and Katelyn

               
Edward F. Pace-Schott, Ph.D.

Director, Sleep and Anxiety Disorders Laboratory
Assistant Professor of Psychiatry
Harvard Medical School
Massachusetts General Hospital - East
CNY 149 13th Street Room 2605
Charlestown, MA 02129
Phone: 508-523-4288
Fax: 617-726-4078
Email: epace-...@mgh.harvard.edu

---

From: Thomas Nichols <thomas....@bdi.ox.ac.uk>
Sent: Thursday, July 30, 2020 6:07 PM
To: Oliver, Katelyn Irene <KIOL...@mgh.harvard.edu>
Cc: Pace-Schott, Edward F.,Ph.D. <EPACE-...@mgh.harvard.edu>; Camprodon, Joan A.,M.D.,M.P.H.,Ph.D. <JCAMP...@mgh.harvard.edu>; snpm-support <snpm-s...@googlegroups.com>; Jeehye Seo <amoi...@gmail.com>; Ellard, Kristen Kathleen <KEL...@mgh.harvard.edu>
Subject: Re: Question