Hi Gwenaëlle,
I think I have got everything you've said about contrasts, but there
is one thing I can't, for the life of me, figure out. How do you set
up an overall F-test. Maybe a step by step instruction could really be
helpful for a newbie like me and maybe for Adnan.
During our first conversation, I said:
> To model a 2x3 interaction with all groups, I used:
> 1/-1/1/-1/1/-1/0/0/0/0/0/0
> -1/1/-1/1/-1/1/0/0/0/0/0/0
You said:
>>Not quite. Not sure here what you're asking for here... but if you want something like >>"overall is there any significant difference between the changes across groups", then you'd >>need an F-test...
Yes, exactly. I think that is what I need, but I don't know how to do it.
I have tried different ways to set it up (using the examples from the
randomize manual as models), but I don't know where to look for the
output or whether or not I did the right setup. Tell me if I am wrong
about this
1. what I am looking for is to see a skelatonized brain image with
lit up regions that show an interaction between the changes between T1
and T2 across the three groups . That is, in the regions that light
up, the differences between T1 and T2 are dependent on the group
(Relaps, abstin, versus control). On SPSS for example this would be a
single line that says something like:
Group by Time/Type III Sum of squares = 150.32/df=6/F=4.2/Sig=.02
How is such output represented in randomize?
2. What would the output be labeled as? When I have tried to do this,
regardless if I click or don't click on the F radio buttons, I get the
same number of outputs. I have put in 7 contrasts, and clicked on
different F radio buttons next to them...that produced 7 contrast
files and nothing else. I am definitely doing something wrong. I have
a major trouble understanding how to use the radio buttons the
represent the F-test.
3. Assuming that my design matrix and group matrix are the ones below,
would you do me a great favor and teach me how to step through the
process of making up a contrast matrix that would give me the F-test
we've been talking about (maybe take baby steps, because I am having
major trouble with this ;) (I am looking for an increase in FA over
time, I don't care about any decreases)
Group/EV1(control time1)/EV2(control time2)/EV3(relapse time 1)/EV4 (relapse
time2)/EV5 (abstinent time1)/EV6 (abstinent time2)/
EV7(subject1)/EV8(sub2)/EV9(sub3)/EV10(sub4)/EV11(sub5)/EV12(sub6)
1/1/0/0/0/0/0/1/0/0/0/0/0
1/0/1/0/0/0/0/1/0/0/0/0/0
2/0/0/1/0/0/0/0/1/0/0/0/0
2/0/0/0/1/0/0/0/1/0/0/0/0
3/0/0/1/0/0/0/0/0/1/0/0/0
3/0/0/0/1/0/0/0/0/1/0/0/0
4/0/0/0/0/1/0/0/0/0/1/0/0
4/0/0/0/0/0/1/0/0/0/1/0/0
5/0/0/0/0/1/0/0/0/0/0/1/0
5/0/0/0/0/0/1/0/0/0/0/1/0
6/0/0/0/0/1/0/0/0/0/0/0/1
6/0/0/0/0/0/1/0/0/0/0/0/1
On Thu, Jan 28, 2010 at 12:54 PM, Gwenaëlle DOUAUD
<[log in to unmask]> wrote:
> Hi Omar,
>
> these two contrasts (for instance):
> 1 -1 -1 1 0 0
> 1 -1 0 0 -1 1
>
> Cheers,
> Gwenaëlle
>
> --- En date de : Jeu 28.1.10, Omar Alhassoon <[log in to unmask]> a écrit :
>
>> De: Omar Alhassoon <[log in to unmask]>
>> Objet: Re: [FSL] Randomize design and contrast matrix for Fractional Anisotropy
>> À: [log in to unmask]
>> Date: Jeudi 28 Janvier 2010, 20h36
>> Thank you Matthew and Gwenaelle,
>>
>> I have been trying to figure how to set up an F-test in Glm
>> that would
>> test the omnibus question of whether there is an
>> interaction between
>> time and group in a 2x3 design. I understand that I need to
>> set up
>> some contrasts, set the F button to 1, then click on the
>> buttons next
>> to the contrasts that I want to include in the F-test. But
>> I have no
>> idea what contrasts should be included in such a design.
>> Any
>> suggestions.
>>
>> Thanks again,
>> Omar
>>
>> On Thu, Jan 28, 2010 at 12:11 PM, Matthew Webster
>> <[log in to unmask]>
>> wrote:
>> > Hello,
>> > If you are using the group column in the GLM to
>> set up a design.grp
>> > file for use with the -e option in randomise, you can
>> safely ignore
>> > this message.
>> >
>> > Many Regards
>> > Matthew
>> >> Thank you for your response, it is very
>> re-assuring.
>> >>
>> >> With regards to the error messages you said:
>> >>
>> >>>It is probably because you've run exactly the
>> model above that is
>> >>>indeed rank deficient. You just need to add
>> some subjects (and
>> >>>presumably you've got more than one control) as
>> some additional EVs and
>> >>>it should not be a problem anymore...
>> >>
>> >> Actually, I have run this model in Glm using 24
>> controls, 15 abstinent
>> >> subjects, and 5 relapsers. That is when the error
>> message appears. So,
>> >> adding the EVs in the same way I did with the
>> example dataset does not
>> >> seem to resolve this error message. Do you have
>> other thoughts about
>> >> why this could be happening?
>> >>
>> >> Thank you so much.
>> >> Omar
>> >>
>> >>
>> >>
>> >>
>> >>> Problem with processing the model. Warning: at
>> least one EV is (close
>> >>> to) a linear combination of the others. You
>> probably need to alter
>> >>> your design. (Design matrix is rank deficient
>> - ratio of min:max
>> >>> eigenvalues in SVD of matrix is 1.21792e-17).
>> >>>
>> >>> Warning-design matrix uses different groups
>> (for different variances),
>> >>> but these do not contain "separable" Evs for
>> the different groups (it
>> >>> is necessary that, for each EV, only one of
>> the groups has non-zero
>> >>> values).
>> >>
>> >> Hope this helps,
>> >> Gwenaelle
>> >>
>> >>
>> >
>>
>
>
>
>
|