Dear Chih-Hao,

If you have a 2x3 factorial design (factor A with two levels and factor
B with 3 levels), you have:
* main effect of A:
  [1 1 1 -1 -1 -1]
* main effect of B:
  [1 -1 0 1 -1 0
   0 1 -1 0 1 -1]
You don't need to include [1 0 -1 1 0 -1] as it is already implicitly
tested by the other contrast weights (a linear combination of them,
their sum here).
* AxB interaction:
  [1 -1 0 -1 1 0
   0 1 -1 0 -1 1]

The SPM function spm_make_contrasts([2 3]) lets you define these
contrasts automatically.

Best regards,
Guillaume.


On 24/01/17 08:53, 連 志浩 wrote:
> Dear Guillaume,
> 
> 
> Thanks for your reply, it's very helpful!
> 
> 
> I had read the technical note, but I have a naive question about F-contrast.
> 
> 
> I wonder why I should input [ 1 -1 0; 0 1 -1] to test a interaction in a
> 2*3 design.
> 
> 
> Couldn't I input [1 0 -1] to test the same interaction?
> 
> 
> What's the difference between [1 -1 0; 0 1 -1] and [1 0 -1] in F-contrast?
> 
> 
> Best regards,
> 
> Chih-Hao
> 
> 
> ------------------------------------------------------------------------
> *寄件者:* Guillaume Flandin <[log in to unmask]>
> *寄件日期:* 2017年1月14日 上午 12:55:10
> *收件者:* 連 志浩
> *副本:* [log in to unmask]
> *主旨:* Re: [SPM] No contrast(s) in flexible factorial design
>  
> Dear Chih-Hao,
> 
> You could test for 1) with a paired t-test.
> 
> 2) and 3) are both testing for a group by condition interaction but they
> will not give you identical results due to different pooled vs
> partitioned error model. This has been discussed at length on this
> mailing list and is described formally in this technical note:
>   http://www.fil.ion.ucl.ac.uk/~wpenny/publications/rik_anova.pdf
> 
> Best regards,
> Guillaume.
> 
> 
> On 12/01/17 08:26, 連 志浩 wrote:
>> Dear Guillaume, 
>> 
>> Thanks for your helps.
>> 
>> I want to make sure that all my contrast weights are right and whether
>> the results from flexible factorial design and two sample t test are same.
>> 
>> 
>> 1. To compare difference between 2 conditions in 1 Group, for example, I
>> want to compare "G1C1 - G1C2".
>> 
>> I guess the contrast weight would be [ 1 -1 0 0 0 0 0 0 zeros(1,18)
>> zeros(1,8) ].
>> 
>> 
>> 2. If I want to know the difference between Groups after both them
>> compared "C1-C2", then the contrast weight should be [ 1 -1 0 0 -1 1 0 0
>> zeros(1,18) zeros(1,8) ].
>> 
>> 
>> 3. If I compare "C1-C2" for each group at 1st-level, then use these con
>> img at two sample t test (2nd-level, Independence: Yes, Variance:
>> Unequal), then input [ 1 -1 ] in vector to compare them.
>> 
>> Would the result from this two sample t test be same with my second
>> question?
>> 
>> 
>> Best regards,
>> 
>> Chih-Hao
>> 
>> 
>> 
>> 
>> ------------------------------------------------------------------------
>> *寄件者:* 連 志浩
>> *寄件日期:* 2017年1月12日 上午 09:23
>> *收件者:* Guillaume Flandin
>> *副本:* [log in to unmask]
>> *主旨:* RE: [SPM] No contrast(s) in flexible factorial design
>>  
>> 
>> Dear Guillaume, 
>> 
>> 
>> Thanks for your help, it's very helpful! 
>> 
>> 
>> I know why I can't do any contrast in flexible factorial design, I
>> thought I don't need to input a value for each column.
>> 
>> 
>> Best regards,
>> 
>> Chih-Hao
>> 
>> ------------------------------------------------------------------------
>> *寄件者:* Guillaume Flandin <[log in to unmask]>
>> *寄件日期:* 2017年1月11日 下午 10:39:50
>> *收件者:* 連 志浩
>> *副本:* [log in to unmask]
>> *主旨:* Re: [SPM] No contrast(s) in flexible factorial design
>>  
>> Dear Chih-Hao,
>> 
>> Thanks - glad to hear you have more subjects! With 18 subjects in group
>> 1 and 8 in group 2, the contrast is:
>>   [1 0 0 0 0 0 0 0 ones(1,18)/18 zeros(1,8)]
>> but you'd be better off doing a one-sample t-test with the 18 images for
>> condition 1 of subjects of group 1.
>> 
>> Best regards,
>> Guillaume.
>> 
>> 
>> On 11/01/17 14:00, 連 志浩 wrote:
>>> Dear Guillaume,
>>> 
>>> 
>>> Thanks for your advice!
>>> 
>>> 
>>> I input more data to run my 2nd-level model, but the same situation
>>> happens again.
>>> 
>>> 
>>> I selected "Specify All" option, and input followed matrix in Factor matrix:
>>> 
>>> [  1    1    1    1
>>>   2    1    1    2
>>>   3    1    1    3
>>>   4    1    1    4
>>>   5    2    1    1
>>>   6    2    1    2
>>>   7    2    1    3
>>>   8    2    1    4
>>> ....
>>> ......
>>>  97   25    2    1
>>>  98   25    2    2
>>>  99   25    2    3
>>> 100   25    2    4
>>> 101   26    2    1
>>> 102   26    2    2
>>> 103   26    2    3
>>> 104   26    2    4 ]
>>> 
>>> In Main effects & Interaction option, I created a main effect (1) and a
>>> interaction (2 3).
>>> 
>>> 
>>> Did I do anything wrong?
>>> 
>>> 
>>> Best regards.
>>> 
>>> Chih-Hao
>>> 
>>> 
>>> ------------------------------------------------------------------------
>>> *寄件者:* Guillaume Flandin <[log in to unmask]>
>>> *寄件日期:* 2017年1月11日 下午 08:50:07
>>> *收件者:* 連 志浩
>>> *副本:* [log in to unmask]
>>> *主旨:* Re: [SPM] No contrast(s) in flexible factorial design
>>>  
>>> Dear Chih-Hao,
>>> 
>>> This is to be expected for a model that includes subject effects. The
>>> corresponding estimable contrast is [1 0 0 0 0 0 1/3 1/3 1/3 0 0 0] but
>>> you are essentially trying to do a one-sample t-test with three subjects
>>> so you should acquire data from more subjects before running a second
>>> level model.
>>> 
>>> Best regards,
>>> Guillaume.
>>> 
>>> 
>>> On 11/01/17 06:55, 連 志浩 wrote:
>>>> Dear experts,
>>>> 
>>>> 
>>>> I have a problem when I tried to use flexible factorial design.
>>>> 
>>>> 
>>>> All the bars below the design matrix are gray, and I can't do any contrast.
>>>> 
>>>> 
>>>> GUI showed "no contrast(s)" when I open SPM contrast manager, and showed
>>>> "!invalid contrast" when I try to input any contrast.
>>>> 
>>>> 
>>>> It's a 2 x 4 flexible factorial design with a Group factor with 2 level
>>>> and a Condition factor with 4 level, and I set 3 factors in my model:
>>>> 
>>>> 
>>>> Factor 1: Subject, Independence: Yes, Variance: Equal
>>>> 
>>>> Factor 2: Group, Independence: Yes, Variance: Unequal
>>>> 
>>>> Factor 3: Condition, Independence: No, Variance: Equal
>>>> 
>>>> 
>>>> I would appreciate for any help.
>>>> 
>>>> 
>>>> Thanks for your reading.
>>>> 
>>>> 
>>>> Best,
>>>> 
>>>> Chih-Hao
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> -- 
>>> Guillaume Flandin, PhD
>>> Wellcome Trust Centre for Neuroimaging
>>> University College London
>>> 12 Queen Square
>>> London WC1N 3BG
>> 
>> -- 
>> Guillaume Flandin, PhD
>> Wellcome Trust Centre for Neuroimaging
>> University College London
>> 12 Queen Square
>> London WC1N 3BG
> 
> -- 
> Guillaume Flandin, PhD
> Wellcome Trust Centre for Neuroimaging
> University College London
> 12 Queen Square
> London WC1N 3BG

-- 
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG