Hi again Tom,
I'm sorry if i didn't make it clear. Yes i have 41 individual patients and
41 individual controls and my data (82 subjects ) were collected as matched
control/patient pairs . My 41 individual patients consist of 23 Siemens
scanned individual patients and another 18 GE scanned individual patients.I
also have 41 individual controls that also consist of 23 Siemens scanned
individuals and 18 GE scanned individuals.
I don't have any previous experience with randomise and am not also so
familiar with high order statistics so forgive me if i'm making silly questions.
In the beginning i was thinking to split my data and analyze them separately
in randomise.
First to keep only the Siemens subjects, make a design matrix according to
the case of unpaired two group difference and then run randomise.
And do exactly the same procedure for the GE data
Do you think that in this way i would have some useful results? Or because
of the fact that i'm splitting my data in two parts i am reducing the power
of the statistics?
In the design that you've proposed and this is the one that i'm gonna
follow,I guess that i should consider only two groups.Right?:
Group Siem GE Cont Pat
1 1 0 1 0
1 1 0 0 1
1 1 0 1 0
1 1 0 0 1
...
2 0 1 1 0
2 0 1 0 1
2 0 1 1 0
2 0 1 0 1
>And contrasts
>Scanner Effect: -1 1 0 0
>Patient/Cont Eff: 0 0 -1 1
>Interaction: -1 1 1 -1
In these 3 contrasts that you've proposed i was thinking of adding another 2
(just the mean of controls and mean of patients so when patient/cont effect
is positive i can see if this is due to patient atrophy or controls
growth).Do u think i should do that?
And one more question...
The patient/control contrast image will give me info where patients exhibit
higher values compared to controls and also the scanner effect contrast
image will give me info where GE scanner exhibits higher values than
Siemens. What kind of information the interaction contrast image will give me?
Finally, the merged data ( 4D file), should follow the previous design
matrix, which means the sequence of the 4D file should be smth like :
Siemens Control
Siemens Patient
Siemens Control
Siemens Patient
...
GE Control
GE Patient
GE Control
GE Patient
Right?
Thank's a lot once more Tom
Antonios-Constantine Thanellas
On Mon, 14 Apr 2008 16:02:40 +0100, Thomas Nichols <[log in to unmask]>
wrote:
>Dear Antonios-Constantine,
>
>I'm afraid I don't quite understand what you're pairing. In stats modelling
>lingo, 'paired' observations usually means two observations from the *same*
>individual. Can you clarify:
>
>How many unique individuals do you have? 41 or 82?
>
>If 41, then the design matrix shown before should work; with annotation
>Group Con:SeiGE Pat:SieGE EV3 ... EV43
> 1 1 0 1 0 0 0 0 0
> 1 -1 0 1 0 0 0 0 0
> 2 1 0 0 1 0 0 0 0
> 2 -1 0 0 1 0 0 0 0
> ...
> 40 0 1 0 0 0 0 1 0
> 40 0 -1 0 0 0 0 1 0
> 41 0 1 0 0 0 0 0 1
> 41 0 -1 0 0 0 0 0 1
>
>And the contrasts would be
>Scanner Eff: 1 1
>Patient/Cont Eff: not estimable
>Interaction: 1 -1
>Note you can't estimate the patient/control main effect because a paired
>analysis discounts any constant effect within a pair as nuisance. However,
>the Scanner-by-Disease interaction is estimable.
>
>If you have 82 unique individuals, then I won't really worry about the
>pairing. If you collected the 82 subjects as matched control/patient pairs,
>that's good, and you can be happy in the knowledge that you've balanced for
>possible confounds; but I won't actually use that information in the
>analysis. I would just treat this as a two-way between subjects ANOVA, with
>a Control/Patient effect, and a Siemens/GE effect, and the interaction.
>
>There are lots of ways to do this, but perhaps the simplest design matrix is
>
>Siem GE Cont Pat
> 1 0 1 0
> 1 0 0 1
> 1 0 1 0
> 1 0 0 1
> ...
> 0 1 1 0
> 0 1 0 1
> 0 1 1 0
> 0 1 0 1
>And contrasts
>Scanner Effect: -1 1 0 0
>Patient/Cont Eff: 0 0 -1 1
>Interaction: -1 1 1 -1
>
>Does this help?
>
>-Tom
>
>
>On Mon, Apr 14, 2008 at 2:44 PM, Antonios - Constantine Thanellas <
>[log in to unmask]> wrote:
>
>> Tom,
>>
>> According to the example that you've gave me my design matrix should look
>> like that:
>>
>> Group EV1 EV2 ..........EV43
>> > 1 1 0 0 1 0 0 0 0 0
>> > 1 -1 0 0 1 0 0 0 0 0
>> > 2 1 0 0 0 1 0 0 0 0
>> > 2 -1 0 0 0 1 0 0 0 0
>> > ...
>> > 40 0 1 0 0 0 0 0 1 0
>> > 40 0 -1 0 0 0 0 0 1 0
>> > 41 0 1 0 0 0 0 0 0 1
>> > 41 0 -1 0 0 0 0 0 0 1
>>
>> Now my groups are not only 2 but 41coupled (since my data set consists of
>> 41
>> pairs of patients devided in 23 Siemens scanned pairs and 18 GE scanned
>> pairs.the same holds for the controls).The way that i interpreted the
>> previous design matrix is the the following:
>>
>> Number of groups :82 (41 couples)
>> Number of EV's: 43
>> The first two EV's express the difference between The Siemens Group
>> (controls-patients) and the GE group (controls-patients)
>> The other 41 Ev's are one for each subject pair (Siemens control-Siemens
>> patient and GE control-GE patient)
>>
>> Group EV1 EV2 ... EV42 EV43
>> 1 1 0 (Siemens Control) 0 0
>> 1 -1 0 (Siemens Patient) 0 0
>> 2 1 0 (Siemens Control) 0 0
>> 2 -1 0 (Siemens Patient) 0 0
>> ...
>> 23 1 0 (Siemens Control) . 0 0
>> 23 -1 0 (Siemens Patient) . 0 0
>> 24 0 1 (GE Control) . 1 0
>> 24 0 -1 (GE Patient) 1 0
>> ...
>> 41 0 1 (GE Control) 0 1
>> 41 0 -1 (GE Patient) 0 1
>>
>> Is this interpretation correct?or did i misunderstand something?
>> If this interpretation holds then i'm quite confused on how it will be
>> possible to extract information concerning the two groups Patients and
>> Controls?Cause instead of having 2 groups now we have 41 and the grouping
>> is
>> among Siemens Scans and GE scans instead of Controls and Patients.Besides
>> this is what EV1 and EV2 express(difference between Siemens group and GE
>> group).Right?
>>
>> In the case of 2 groups (patients and controls), you just need 4 contrasts
>> (A-B,B-A,A,B) to extract this info but how about my case?
>>
>> I would really appreciate your help once more cause i'm stuck here.
>> Thank you again
>> Antonios-Constantine Thanellas
>>
>>
>>
>>
>> On Fri, 11 Apr 2008 11:55:53 +0100, Thomas Nichols <[log in to unmask]
>> >
>> wrote:
>>
>> >Antonios,
>> >
>> >This design would be a 'two group, paired difference'. Your design
>> matrix
>> >should look like the Feat " (Two-Sample Paired T-Test)" design *but* with
>> >two columns for the paired differences.
>> >
>> >Using a slightly different order than the Feat web page (i.e. grouping
>> >subjects together) the design matrix will look like
>> > 1 0 0 1 0 0 0 0 0
>> >-1 0 0 1 0 0 0 0 0
>> > 1 0 0 0 1 0 0 0 0
>> >-1 0 0 0 1 0 0 0 0
>> >...
>> > 0 1 0 0 0 0 0 1 0
>> > 0 -1 0 0 0 0 0 1 0
>> > 0 1 0 0 0 0 0 0 1
>> > 0 -1 0 0 0 0 0 0 1
>> >but then you *also* need to define a group file which tells randomise
>> about
>> >the pairing
>> >1
>> >1
>> >2
>> >2
>> >3
>> >3
>> >...
>> >That should do it.
>> >
>> >-Tom
>> >
>> >
>> >On Thu, Apr 10, 2008 at 3:43 PM, Antonios - Constantine Thanellas <
>> >[log in to unmask]> wrote:
>> >
>> >> Dear fsl users,
>> >>
>> >> I have 30 T1 pairs (baseline and follow up) of controls scanned with
>> >> Siemens
>> >> scanner and 21 T1 pairs scanned with GE scanner
>> >>
>> >> My patients consist of 23 pairs of Siemens scanner and 18 pairs of GE
>> >> scanner
>> >>
>> >> After longitudinal analysis i want to continue with the use of
>> randomise
>> >> and
>> >> localize differences between controls and patients.
>> >>
>> >> I supose that the first step is to use grouped matched controls and
>> >> patients
>> >> (23 Siemens and 18 GE controls as one group and the second group 23
>> >> Siemens
>> >> and 18 GE patients). Is my assumption correct?
>> >>
>> >> I'm quite confused on how am i gonna set up the contrasts and the
>> design
>> >> matrix. Should i consider my case as an Unpaired Two-Group Difference
>> >> (Two-Sample Unpaired T-Test) and set up my contrast files as it is
>> >> mentioned
>> >> in the corresponding part of the help page of FEAT or my case falls
>> into
>> >> the
>> >> case of Paired Two-Group Difference (Two-Sample Paired T-Test)?
>> >>
>> >> Thanks in advance
>> >> Antonios-Constantine Thanellas
>> >>
>> >>
>> >
>> >
>> >--
>> >____________________________________________
>> >Thomas Nichols, PhD
>> >Director, Modelling & Genetics
>> >GlaxoSmithKline Clinical Imaging Centre
>> >
>> >Senior Research Fellow
>> >Oxford University FMRIB Centre
>> >
>>
>>
>>
>>
>
>
>--
>____________________________________________
>Thomas Nichols, PhD
>Director, Modelling & Genetics
>GlaxoSmithKline Clinical Imaging Centre
>
>Senior Research Fellow
>Oxford University FMRIB Centre
>
|