I'm not sure. I've always use the "Specify All" option in the flexible
factorial model as I never had the patience to enter each pair
manually.
More recently, I've switched to GLM Flex for all my analysis as its
very easy to add subjects to the existing scripts.
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 Fri, Nov 9, 2012 at 2:12 AM, Zhenhao Shi <[log in to unmask]> wrote:
> Hi Donald and Chao and all SPMers,
>
> Thank you very much for your reply. I guess I found what went wrong, and it
> seems to be SPM's fault but not mine.
>
> I rechecked my batch to see if I had paired the con*.img files wrong, and
> they were all paired correctedly as con_0001.img & con_0002.img for each
> subject. I then estimated it again and reviewed the design matrix, which was
> like
> 1 0 1 0 0 0 0 0 0 0 0 0
> 0 1 1 0 0 0 0 0 0 0 0 0
> 1 0 0 1 0 0 0 0 0 0 0 0
> 0 1 0 1 0 0 0 0 0 0 0 0
> 1 0 0 0 1 0 0 0 0 0 0 0
> 0 1 0 0 1 0 0 0 0 0 0 0
> 1 0 0 0 0 1 0 0 0 0 0 0
> 0 1 0 0 0 1 0 0 0 0 0 0
> ...
> such that the first two columns correspond to con_0001 and con_0002, and the
> rest ones are subjects. However, when I checked "Files and factors", it was
> like
> #img sF3 sF2 filename tails
> 001 01 01 Sub01_con_0001.img,1
> 002 02 01 Sub02_con_0001.img,1
> 003 01 02 Sub03_con_0001.img,1
> 004 02 02 Sub04_con_0001.img,1
> 005 01 03 Sub05_con_0001.img,1
> 006 02 03 Sub06_con_0001.img,1
> 007 01 04 Sub07_con_0001.img,1
> 008 02 04 Sub08_con_0001.img,1
> ...
> 020 02 10 Sub20_con_0001.img,1
> 021 01 11 Sub01_con_0002.img,1
> 022 02 11 Sub02_con_0002.img,1
> 023 01 12 Sub03_con_0002.img,1
> 024 02 12 Sub04_con_0002.img,1
> ...
> I think #img corresponds to the line number of the design matrix, sF3
> corresponds to con_0001/con_0002, and sF2 is the subject number. If I'm
> right about this, then what SPM did is totally wrong, because it made me
> compare the wrong con*.img's using [1 -1].
>
> As Donald suggested, I opened SPM.mat and looked into SPM.xY.P, which is a
> 20x2 cell, the first column containing con_0001.img files and the second
> containing con_0002.img files. I then looked into SPM.xY.VY.fname, where
> SPM.xY.VY(1:20).fname are con_0001.img's, SPM.xY.VY(21:40).fname are
> con_0002.img's, which are also incorrect.
>
> I haven't updated SPM for a while. Hoping this bug has disappeared in later
> versions and not affected many people's analysis. Or, if I did some steps
> wrong to cause this glitch, please help me out.
>
> Thanks a lot!
>
> Best,
> Zhenhao
>
>
>
> -----
>
> Zhenhao SHI 石振昊
>
> Culture and Social Cognitive Neuroscience Lab
> Department of Psychology
> Peking University
> 5 Yiheyuan Road
> Beijing 100871, P.R.China
> Phone: 86 134 6655 0474
> Email: [log in to unmask]
> http://www.psy.pku.edu.cn/LABS/CSCN_lab
>
>
>
> On Fri, Nov 9, 2012 at 2:09 PM, Chao Suo <[log in to unmask]> wrote:
>>
>> Dear Zhenhao,
>>
>> I asked myself the similar question before.
>>
>> I think it is wrong by saying “the paired t-test on A>Neutral and
>> B>Neutral should be identical to the one-sample t-test on A>B, right? “. I
>> guess you assume that One way t-test is “calculate the average”; and paired
>> t-test is not “minus each other”, which is not that simple.
>>
>>
>>
>> The contrast you put in is calculating the beta values which is means how
>> well this activation(intensity) of this voxel fits a particular factor in
>> your GLM design. It means your beta values will be different if you use two
>> different second level design.
>>
>>
>>
>> BTW, “MriCroN, I see that con_0003.img and con_1vs2.img have the same
>> voxel values” is correct, as they are in the same GLM design. So you can get
>> this by calculating the contrast (suppose your order is A B N) : [1 0 -1]-[0
>> 1 -1]=[1 -1 0] which is A>B.
>>
>>
>>
>> Hope this help.
>>
>>
>>
>> Cheers
>>
>>
>>
>> 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 | Web: rng.org.au/chao-suo-2/
>>
>>
>>
>> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
>> Behalf Of Zhenhao Shi
>> Sent: Friday, 9 November 2012 12:49 AM
>> To: [log in to unmask]
>> Subject: [SPM] One-sample t-test versus paired t-test
>>
>>
>>
>> Dear SPMers,
>>
>>
>>
>> I have a question regarding the difference between one-sample t-test and
>> paired t-test in SPM8.
>>
>>
>>
>> I got one single group of 20 subjects with Condition A, Condition B, and a
>> Neutral baseline, and I would like to compare A and B. In the first level, I
>> did three T-contrasts for each subject: "A > Neutral", "B > Neutral", and "A
>> > B", which resulted in three contrast images: con_0001.img, con_0002.img,
>> con_0003.img. In the second level, I did a one-sample t-test using the files
>> of con_0003.img from all subjects, and entered [1] for a t-contrast, which
>> gave me reasonable brain activations for A > B. After that, just out of
>> curiosity, I tried paired t-test and entered con_0001.img and con_0002.img
>> in pairs each subject at a time. The design matrix looks like that the first
>> two columns are of my interest and the next 20 columns are for each subject.
>> I then entered [1 -1] for a t-contrast, but obtained totally different
>> results from those in the previous one-sample t-test. From my understanding,
>> the paired t-test on A>Neutral and B>Neutral should be identical to the
>> one-sample t-test on A>B, right?
>>
>>
>>
>> To figure out what went wrong, I tried using spm_imcalc_ui to manually
>> substract con_0002.img from con_0001.img for each subject, and got another
>> 20 new image files I named as con_1vs2.img. Using MriCroN, I see that
>> con_0003.img and con_1vs2.img have the same voxel values. So, it looks as if
>> my first level analysis was right. Then could the different results of
>> one-sample t-test and paired t-test be caused by their different algorithms?
>> Or there must be some mistakes I managed not to notice?
>>
>>
>>
>> Looking forward to you experts' reply. Thanks in advance!
>>
>>
>>
>> Best,
>>
>> Z
>>
>>
>>
>>
>> -----
>>
>> Zhenhao SHI
>> Culture and Social Cognitive Neuroscience Lab
>> Department of Psychology
>> Peking University
>> 5 Yiheyuan Road
>> Beijing 100871, P.R.China
>> Email: [log in to unmask]
>> http://www.psy.pku.edu.cn/LABS/CSCN_lab
>
>
|