Print

Print

Thank you very much, Donald. The script went smoothly and I opened the result image in orthoview without any problems.

My matlab version is 7.8.0.347 (R2009a), working on a Vista 64-bit PC.

 

I have a few further questions:

 

1, in the test of the 3-way interaction, what is the actually contrast?

Is it [-1 1 1 -1 1 -1 -1 1] with zeros on others?

I.Cons(6).name = 'P_ by C_ by Time';

I.Cons(6). Groups = {15 16 17 18 19 20 21 22};

I.Cons(6).Levs = [2 2 2];

I.Cons(6).ET = [];

I.Cons(6).mean = 0;

2,Our hypothesis is that the

combination effect of the increase > effect X with Y off > none effect.

Or, combination effect of the increase > effect Y with X off> none effect

If the it is Yes in Q1, I don’t think I am testing the interaction that I need. In fact, I did some post hoc analysis, and they didn’t show these gradient trends.

 Should it be an f contrast to do the job?

For example of “combination effect of the increase > effect X with Y off > none effect.”, this contrast is a combination of two t-contrasts:

Zeros (1,6+8) [-1 1 0 0 1 -1 0 0] [0…0]    % XY>X

Zeros (1,6+8) [0 0 0 0 -1 1 1 -1] [0…0]     %X>None

 

If it is correct, how can I “AND” these two contrasts? Simply By overlapping the two result images?

 

The design is [effect X(x1,x2), effect Y (y1,y2), time(t1,t2), (x1t1,x1t2,x2t1,x2t2,y1t1,y1t2,y2t1,y2t2), (x1y1t1,x1y1t2,x1y2t1,x1y2t2,x2y1t1,x2y1t2,x2y2t1,x2y2t2), subjects (n1+n2+n3+n4)]

Please note that (1=on, 2=off).

 

3, Can I actually design my own contrast on GLM_Flex like in the SPM?

I understand I can easily do this in SPM, however, SPM doesn’t allow me to put a three way interaction in the Flex-Factorial design matrix.

So I have to use GLM_Flex to build it. What should I do if I want to define some new contrast in the design matrix built by GLM_Flex?

 

4, Does the MC FDR in Orthroview means cluster level?

 

5, I the mail-list you mentioned that Orthoview presents both negative and positive results, and post hoc analysis are needed if I want to know which direction they are.

I found it a bit time consuming. What I did is to extract the raw values from the significant clusters in Matlab and test them on SPSS. Is there any easier way to do it?

 

Thanks

 

Chao

 

 

From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of MCLAREN, Donald
Sent: Saturday, 19 May 2012 1:57 AM
To: [log in to unmask]
Subject: Re: [SPM] 2x3x3 ANOVA (within-subject)

 

We are in the process of updating the some of the MATLAB code. The current code was developed for and has been tested in MATLAB 2010. Thus, there are no guarantees of backwards compatibility. However, there are some easy fixes to make in compatible with older versions.

The error related to EDF field missing can be fixed by changing these lines in estimateError.m:

RES = [];
SS = [];

(both of which are around line# 68) to

RES = {};
SS = {};

Let me know if you have additionally issues. In the future, please provide the version of MATLAB that you are using and the OS platform.

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
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 Thu, May 17, 2012 at 8:02 PM, Chao Suo <[log in to unmask]> wrote:

Dear Donald and SPMers,

 

I have figured out that GLM_Flex can do this analysis. My design is 2*2*2, two 2-level between group factor, one 2-level within subject factor (time). According to the example online, I set up my GLM_Flex as below:

 

 

IN.N_subs = [ 17 15 20 21]

IN.Between=[2 2]

IN.BetweenLabs={{'P_On', 'P_Off'} {'C_On','C_Off'}}

IN.Within=[2]

IN.WithinLabs={{'BL' 'FU1'}}

IN.Interactions={[1 3] [2 3] [1 2 3]}

In.FactorLabs={'P_*Time' 'C_*Time' 'P_*C_*Time'}

IN.FactorLabs={'P_*Time' 'C_*Time' 'P_*C_*Time'}

IN.EqualVar=[0 0 1]

IN.Independent=[1 1 0]

F=CreateDesign(IN)

% figure(20); imagesc(F.XX); shg

 

I.OutputDir = pwd;

I.Scans = AllScans;

I.F = F;

I.RemoveOutliers = 0;

I.minN = 2;

I.DoOnlyAll = 1;

I.CompOpt=0;

 

I=GLM_Flex(I);

 

I.Cons(1).name = 'P_';

I.Cons(1).Groups = {1 2};

I.Cons(1).Levs = [2];

I.Cons(1).ET = [];

I.Cons(1).mean = 0;

 

I.Cons(2).name = 'C_';

I.Cons(2).Groups = {3 4};

I.Cons(2).Levs = [2];

I.Cons(2).ET = [];

I.Cons(2).mean = 0;

 

I.Cons(3).name = 'Time';

I.Cons(3).Groups = {5 6};

I.Cons(3).Levs = [2];

I.Cons(3).ET = [];

I.Cons(3).mean = 0;

 

I.Cons(4).name = 'P_ by Time';

I.Cons(4).Groups = {7 8 9 10};

I.Cons(4).Levs = [2 2];

I.Cons(4).ET = [];

I.Cons(4).mean = 0;

 

I.Cons(5).name = 'C_ by Time';

I.Cons(5).Groups = {11 12 13 14};

I.Cons(5).Levs = [2 2];

I.Cons(5).ET = [];

I.Cons(5).mean = 0;

 

I.Cons(6).name = 'P_ by C_ by Time';

I.Cons(6). Groups = {15 16 17 18 19 20 21 22};

I.Cons(6).Levs = [2 2 2];

I.Cons(6).ET = [];

I.Cons(6).mean = 0;

 

I=GLM_Flex_Contrasts(I)

 

 

The matrix was fine (attached below). However, I got the error message:

Writing out mat files and image files...

 

Using ResMS #1

??? Reference to non-existent field 'EDF'.

 

Error in ==> GLM_Flex_Contrasts at 98

            df2 = oo.EDF{1}(errLev);

 

Error in ==> MY_ALL_GLM at 208

I=GLM_Flex_Contrasts(I)

 

There are only generate two ResMS_ nii files. And they are all invalid to be open.

I traced back a bit about this error, and I found there is no EDF field in oo.mat file.

Is it because of some error in my design?

 

Thanks

 

Chao

 

 

-----Original Message-----
From: Chao Suo
Sent: Tuesday, 15 May 2012 1:17 PM
To: [log in to unmask]
Cc: Chao Suo
Subject: Can I test the interaction of three factors in flexible factorial design

 

Dear SPMers,

 

My design matrix of these structural images includes four factors:

1,subject

2,effect X

3,effect Y

4, time.

 

In the matrix, we want to put four main effects (subject, effect X, effect Y, time),interactions of (effect X*time), (effect Y*time) and (effect X*effect Y*time). However, it seems that the the input of interaction is set to be a 1by2 vector. Does that mean that SPM couldn't cope with this interaction of three (or more) factors?

 

Thanks for your help.

 

Chao Suo,

PhD Candidate, School of Psychiatry, UNSW, Australia

Brain & Mind Research Institute THE UNIVERSITY OF SYDNEY

Room 401, Level 4, M02K | 100 Mallett St Camperdown | NSW | 2050

T +61 2 9351 0728  | F +61 2 9351 0551

 

 

 

From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of MCLAREN, Donald
Sent: Sunday, 26 February 2012 5:06 AM
To: [log in to unmask]
Subject: Re: [SPM] 2x3x3 ANOVA (within-subject)

 

GLM_Flex will handle your design without a problem.


Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
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 Sat, Feb 25, 2012 at 12:49 PM, Gabor Oederland <[log in to unmask]> wrote:

Dear all,


I ran into troubles with a 2x3x3-Within-Subject-Design. The idea is one group with 14 subjects, three factors, overall this makes a 2x3x3-Design (I ran models on single-subject level and now would like to conduct a group analysis based on the con images. Due to the Design, I have 18 con images per subject, each con images corresponding to one stimulus combination and contrasted against implicit baseline).

I'm interested in any of the potential main effects as well as the three possible two-way interactions and the three-way interaction.

"Flexible factorial" does not allow for any three-way interactions, so this option seems to be useless. The extension GLM_flex http://nmr.mgh.harvard.edu/harvardagingbrain/People/AaronSchultz/GLM_Flex.html does not provide any solution for my design as well.

Thus it seems to me that I have to run a "Full Factorial" as it is the only design in which I can look at three-way interactions at all. So I set up a model with three factors with "Independence - No" and "Variance - Equal". But as far as I understood based on other messages dealing with 2x2 or 2x3 ANOVAs, this is statistically wrong for a within-subject design?

Would it help to add another factor "Subject"? If not, any other ideas?


Your help is greatly appreciated!

Gabor