Hi,
On 24 Jun 2007, at 14:08, Remya Nair wrote:
> Hi Steve-
>
> I have looked over my original email and you were right to be
> confused as it wasn't clear.
> Let me restate the situation:
>
> 2 groups (groupA, groupB) with 12 subjects each.
> each subject did 8 runs (4 for condition C, 4 for condition P)
> the lower level analysis was done for each run with 3EVs (ML, MH, SH)
> and 4 contrasts:
> 1 0 0
> 0 1 0
> 0 0 1
> 0 1-1
> then a higher level analysis was done to combine the 4 runs of each
> condition for each subject.
> this leaves us with 2 gfeat directories for each subject - one for
> condition C and one for condition P. each has four cope.feat
> directories corresponding to ML, MH, SH and MH-SH from the lower-
> level analyses.
Right - you mean you did a separate second-level analysis for each
subject.
> say what we want to compare:
> groupA vs groupB, ML - MH for condition C
> ML is cope1, and MH is cope2, from the gfeat directories.
>
> therefore for the higher-level analysis we have 48 inputs:
> first 12 are groupA condition C, cope1 (EV1)
> next 12 are groupA condition C, cope2 (EV2)
> then 12 groupB, condition C, cope1 (EV1)
> then 12 groupB, condition C, cope2 (EV2)
>
> would the contrast be [1 -1 -1 1]?
Yes - the contrast is correct but the degrees of freedom isn't (you
"appear" to have 48 independent subjects instead of 24 at this
highest level - hence you will get an inflated false positive rate.
There are two ways of fixing this:
- At the first or second level, create the ML-MH contrast so that you
only have to feed in 24 inputs to the 3rd level.
OR
- Setup the 3rd level as a "paired t-test" equivalent - i.e. add a
new EV to model each subject's mean of ML and MH (see the manual for
examples).
Cheers.
>
> please let me know if this still isn't clear.
> thanks again!
> R
>
>
> On Jun 23, 2007, at 2:13 AM, Steve Smith wrote:
>
>> Hi, let's reconfirm what's in this higher-level analysis:
>>
>> In this particular analysis, all entries correspond to ML-MH
>>
>> The first 12 entries (modelled by EV1) are groupA condition C
>> The next 12 (EV2) are groupA condition P
>> Then group B condition C
>> Then group B condition P
>>
>> Is this not correct? I wasn't 100% sure from your original email
>> what was in this higher-level analysis.
>>
>> Cheers.
>>
>>
>>
>> On 21 Jun 2007, at 18:57, Remya Nair wrote:
>>
>>> Hi Steve-
>>>
>>> Thanks for your help.
>>> I just want to confirm one thing though...
>>>
>>> I understand with the way you set up the EVs, but to compare
>>> groupA vs groupB ML-MH for condition C, shouldn't the contrast be
>>> [1 -1 -1 1]?
>>> and not [1 0 -1 0] as you wrote.
>>> My understanding is that would be a comparison of groupA cope1 vs
>>> groupB cope1, right?
>>>
>>> Thanks again.
>>> R
>>>
>>> On Jun 17, 2007, at 6:08 PM, Steve Smith wrote:
>>>
>>>> Hi,
>>>>
>>>> On 15 Jun 2007, at 17:14, Remya Nair wrote:
>>>>
>>>>> hi-
>>>>>
>>>>> i was hoping someone could help me set up an anova.
>>>>>
>>>>> let me describe the experiment:
>>>>>
>>>>> there are 2 groups with 12 subjects in each group.
>>>>> each subject did 8 runs (4 run for condition C and 4 runs for
>>>>> condition P, with 3 levels for each condition (ML/MH/SH)).
>>>>>
>>>>> i ran the first-level analysis with 3EVs (ML/MH/SH) and then
>>>>> combined
>>>>> the 4 runs of a single condition.
>>>>>
>>>>> now i have 2 .gfeat directories for each subject: 1 for
>>>>> condition C
>>>>> and one for condition P (each has 4 cope.feat dirs - ML, MH, SH,
>>>>> MH-SH).
>>>>>
>>>>> i want to look at the group differences across different
>>>>> conditions
>>>>> e.g., groupA vs groupB, looking at ML-MH for condition C.
>>>>>
>>>>> so i would have 48 inputs in my higher level analysis:
>>>>> 12 groupA cope1
>>>>> 12 groupA cope2
>>>>> 12 groupB cope1
>>>>> 12 groupB cope2
>>>>>
>>>>> but i am not sure how to set up my EVs and contrasts.
>>>>
>>>> Something like:
>>>>
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 1 0 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 1 0 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 1 0
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>> 0 0 0 1
>>>>
>>>> and for the question:
>>>>> e.g., groupA vs groupB, looking at ML-MH for condition C.
>>>> use the contrast [ 1 0 -1 0] etc.
>>>>
>>>>> also, what if i want to look at ML vs. (MH-SH)?
>>>> Probably the best thing is to achieve this contrast at first-
>>>> level making the higher-level analyses easy. At first level this
>>>> would be:
>>>> [1 -1 1] though I'm not sure whether you do exactly mean ML -
>>>> (MH-SH)
>>>>
>>>>> likewise, is it possible to look at group differences across
>>>>> conditions C+P?
>>>>
>>>> Sure - in a similar manner to the above.
>>>>
>>>> Cheers, Steve.
>>>>
>>>>
>>>>
>>>>>
>>>>> thank you in advance!
>>>>>
>>>>> R
>>>>
>>>>
>>>> -------------------------------------------------------------------
>>>> --------
>>>> Stephen M. Smith, Professor of Biomedical Engineering
>>>> Associate Director, Oxford University FMRIB Centre
>>>>
>>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>>> +44 (0) 1865 222726 (fax 222717)
>>>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>>> -------------------------------------------------------------------
>>>> --------
>>>>
>>
>>
>> ---------------------------------------------------------------------
>> ------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------
>> ------
>>
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|