On Thu, Dec 6, 2012 at 2:11 AM, Qasim Bukhari <[log in to unmask]> wrote:
> Thanks a lot for your reply.
> So then it is group fixed analysis.
> Regarding the null hypothesis, this is exactly what my experiment#4 does.
> And surely as expected there appears a brain network that corresponds to the
> seed region connectivity. However what I want to test is; also "PROVE" the
> null hypothesis rather than "DISPROVING" it; just to check it data is all
> good and there are no undesired connectivity appearing for any reason. The
> experiment that you describe surely can disprove the null hypothesis, which
> is normally what we do; however; can you please also suggest how to prove
> the null hypothesis ? Please find below, how I think I can do it, and you
> may correct me if I m wrong. There are two ways I can think of, it can be
> done.

You cannot prove the null hypothesis, you can only reject it.


>
> Exp 1: Taking subjects from different groups
> Step 1.  The null hypothesis is the the connectivity in group 1 equals the
> connectivity in group 2. Ho: G1= - G2
> Step 2: Convert this to a contrast 1 over the G1 column and 1 over the G2
> column.
> Step 3: I would have had expected NO CONNECTIVITY, but unfortunately I still
> see connectivity.
> Have I done something wrong, or misunderstood something ?

I would expect to see a lot of connectivity as very few areas will
have G1=-G2. It seems that you want to know that the connectivity
exists in each group before testing group differences. You should only
put in one group at a time and then test for connectivity. You will
find the DMN if you use a DMN seed.


>
>
> Exp2:Taking subjects from same group, that means they dont have any
> difference.
> Step 1.  The null hypothesis is the the connectivity in group 1 equals the
> connectivity in group 2. Ho: G1=  G2.
> Step 2: Convert this to a contrast 1 over the G1 column and -1 over the G2
> column.
> Step 3: I would have had expected NO CONNECTIVITY, but unfortunately I still
> see connectivity.
> This was how you described also, but just to see to "prove" the null
> hypothesis, I took the samples from the "same" group, however I still see
> the connectivity. Have I done something wrong, or misunderstood something ?

You are finding where your two subjects are different. This is a fixed
effects analysis with many degrees of freedom.

>
> What could be an experiment showing "NO connectivity" using "group fixed
> effect" when the subjects are drawn from the same group; and showing
> "connectivity" when they are drawn from different group; provided we are
> doing "group fixed effect" on resting state ? This is all I want to do, but
> for some reason I see connectivity in all cases.

I am not sure. Whenever I do resting state analysis, I check each
subject. If the subject has connectivity, then it goes into a group
random effects analysis. I never do a fixed effects analysis.

>
> Best regards,
> Qasim
>
>
>
> Qasim Bukhari
>
> Research Assistant and Doctoral Candidate
>
> Institute for Biomedical Engineering
>
> ETH and University Zurich
>
> Wolfgang-Pauli-Strasse 27, HIT E22
>
> webpage: http://www.micro.biol.ethz.ch/people/sybukhar/index
>
>
>
>
>> Date: Wed, 5 Dec 2012 19:36:02 -0500
>> From: [log in to unmask]
>
>> Subject: Re: [SPM] 1st level analysis on F test
>> To: [log in to unmask]
>
>>
>> On Wed, Dec 5, 2012 at 9:37 AM, Qasim Bukhari <[log in to unmask]>
>> wrote:
>> > I m not testing group differences in 1st level, I want to test fixed
>> > effect.
>> > For fixed effect analysis, there is no second level. And I intend to go
>> > towards second level/random effect, once my fixed effect analysis is
>> > true.
>>
>> >>> I would call this a group fixed analysis to avoid confusion. Calling
>> >>> it a first-level implies subjects AND implies that there would be a second
>> >>> level.
>>
>> > Regarding null hypothesis, how would I define is then, and where exactly
>> > I
>> > have to define it ? According to my understanding, null hypothesis is
>> > defining that there exists no effect; how can I define it in my case
>> > then?
>> > during 'fMRI model specification' ?
>> > P.S. I m working with resting state data, so there is no 'conditions'
>> > that
>> > are defined.
>>
>>
>> The null hypothesis is the the connectivity in group 1 equals the
>> connectivity in group 2.
>> Ho: G1=G2
>>
>> Step 1: Make the null hypothesis equal to 0.
>> Ho: G1-G2=0
>>
>> Step 2: Convert this to a contrast
>> 1 over the G1 column and -1 over the G2 column, all other columns
>> should be 0 (if they exist)
>>
>> ==
>> The F-contrast would be testing that the sum/average of G1 and G2
>> equals 0. I would expect you would get one brain network appearing
>> that corresponds to the seed region.
>>
>> Hope this helps.
>>
>> >
>> >
>> > Qasim Bukhari
>> >
>> > Research Assistant and Doctoral Candidate
>> >
>> > Institute for Biomedical Engineering
>> >
>> > ETH and University Zurich
>> >
>> > Wolfgang-Pauli-Strasse 27, HIT E22
>> >
>> > webpage: http://www.micro.biol.ethz.ch/people/sybukhar/index
>> >
>> >
>> >
>> >
>> >> Date: Wed, 5 Dec 2012 09:20:20 -0500
>> >> Subject: Re: [SPM] 1st level analysis on F test
>> >> From: [log in to unmask]
>> >> To: [log in to unmask]
>> >> CC: [log in to unmask]
>> >
>> >>
>> >> The null hypothesis is defined before you create the contrasts. The
>> >> contrasts are based on your null hypothesis.
>> >>
>> >> You should not be testing group differences in a 1st level model;
>> >> rather, you should take estimates of the first level model from each
>> >> subject to the second level random effects model.
>> >>
>> >> 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 Wed, Dec 5, 2012 at 9:02 AM, Qasim Bukhari <[log in to unmask]>
>> >> wrote:
>> >> > Dear SPM Expert,
>> >> > I m trying to run an analysis differentiating between the two
>> >> > population. I
>> >> > have two groups, group 1 and group 2. These two group differ under
>> >> > one
>> >> > of
>> >> > the regressor; which is an extracted time series of region R1. I
>> >> > tried
>> >> > to
>> >> > run 3 different experiments
>> >> >
>> >> > Experiment 1: group 1 vs group 1 : fixed effect analysis
>> >> > I input two different subjects but both from group 1. Under the
>> >> > multiple
>> >> > regressors; I input the extracted time series of region R1. In order
>> >> > to
>> >> > see
>> >> > the results; I defined an F-contrast, with (1,1) imagining that its
>> >> > the
>> >> > same
>> >> > group so the contrast should also be the same. However in the result;
>> >> > I
>> >> > see
>> >> > quite a lot activations. I wasnt expecting that, since these are from
>> >> > the
>> >> > same population
>> >> >
>> >> > Experiment 2: group 2 vs group 2 : fixed effect analysis
>> >> > I did the same procedure as described above however this time for
>> >> > group
>> >> > 2.
>> >> > The results were same as mentioned above; while I was expecting
>> >> > contrary.
>> >> >
>> >> > Experiment 3: group 1 vs group 2 : fixed effect analysis
>> >> > This time, I input different subjects from the different groups;
>> >> > however
>> >> > I
>> >> > kept the F-contrast as 1, 1. Again I see a lot of activations in the
>> >> > result,
>> >> > and I dont have any explanation for this actually
>> >> >
>> >> > Experiment 4: group 1 vs group 2 : fixed effect analysis
>> >> > Same experiment as experiment 3; however I changed the F contrasts to
>> >> > 1,
>> >> > -1.
>> >> > Once again I see a lot of activations. Precisely the regressor is
>> >> > region
>> >> > R1;
>> >> > which is the acting differently between group 1 and group 2; and
>> >> > while
>> >> > doing
>> >> > fixed effect from different population, I would have had expected the
>> >> > activations in experiment 4. But I m not able to interpret the
>> >> > results
>> >> > from
>> >> > other 3 experiments then. Have I understood something wrong ??
>> >> >
>> >> > I have another question; when can I define my null hypothesis; can I
>> >> > define
>> >> > it before the 2nd level analysis ?? If I understand correctly my null
>> >> > hypothesis should be defined with the contrast (1,1) ?? is it correct
>> >> > ??
>> >> >
>> >> > Thanks a lot
>> >> > best regards,
>> >> > Qasim
>> >> >
>> >> >
>> >> >
>> >> > Qasim Bukhari
>> >> >
>> >> > Research Assistant and Doctoral Candidate
>> >> >
>> >> > Institute for Biomedical Engineering
>> >> >
>> >> > ETH and University Zurich
>> >> >
>> >> > Wolfgang-Pauli-Strasse 27, HIT E22
>> >> >
>> >> > webpage: http://www.micro.biol.ethz.ch/people/sybukhar/index
>> >> >
>> >> >