I see, this makes sense. Two brief questions to (hopefully) conclude:
1. With this model, I can test pairwise contrasts but cannot test
columns by themselves (i.e., each condition against implicit
baseline), is this correct?
2. I can test conjunction nulls as usual simply by selecting multiple
pairwise comparisons, as in the attached image?
Thanks very much for sharing your expertise.
On Tue, Sep 13, 2011 at 3:02 PM, MCLAREN, Donald
<[log in to unmask]> wrote:
> Yes. It is rank-deficient. This is expected and is fine. Its an
> expansion of the paired T-test, which is also rank-deficient.
>
> All it is saying is that the columns are dependent on each other,
> which we know is the case since the the conditions come from the same
> subject.
>
> Best Regards, Donald McLaren
> =================
> D.G. McLaren, Ph.D.
> Postdoctoral Research Fellow, GRECC, Bedford VA
> Research Fellow, Department of Neurology, Massachusetts General Hospital and
> Harvard Medical School
> 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 Tue, Sep 13, 2011 at 5:57 PM, Bob Spunt <[log in to unmask]> wrote:
>> When adding the main effect for subject, the design matrix is rank
>> deficient (see attached "design_wsubme.png"). Thoughts on how I should
>> move forward? Thanks for any tips.
>>
>> On Tue, Sep 13, 2011 at 2:41 PM, MCLAREN, Donald
>> <[log in to unmask]> wrote:
>>> You need to add a main effect for Subject.
>>>
>>> You should also change the variance to equal for subject.
>>>
>>> Best Regards, Donald McLaren
>>> =================
>>> D.G. McLaren, Ph.D.
>>> Postdoctoral Research Fellow, GRECC, Bedford VA
>>> Research Fellow, Department of Neurology, Massachusetts General Hospital and
>>> Harvard Medical School
>>> 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 Tue, Sep 13, 2011 at 3:39 PM, Bob Spunt <[log in to unmask]> wrote:
>>>> Thank you both. I have some follow-up questions to clarify. When I run
>>>> the original configuration the resulting design matrix (see attached
>>>> "design.png") only contains columns for the four conditions. Is it
>>>> invalid to use this model to test the conjunction null using the
>>>> contrasts depicted in attached "contrast.png"? If so, why is it
>>>> invalid?
>>>>
>>>> Donald: when you say "subject...needs to be in the model", do you mean
>>>> that I need to add the main effect of subject, or is the way I had it
>>>> originally set up sufficient (where the subject factor is explicitly
>>>> defined and correctly specified in the factor matrix). If the former,
>>>> it's unclear to me how to add the main effect of subject without
>>>> making the design rank deficient.
>>>>
>>>> Thanks again for your help.
>>>>
>>>> On Tue, Sep 13, 2011 at 9:04 AM, MCLAREN, Donald
>>>> <[log in to unmask]> wrote:
>>>>> (1) I think subject should be set to equal variance AND it needs to be
>>>>> in the model.
>>>>>
>>>>> (2) Some of the conjunctions are impossible, some are possible, here is why:
>>>>> A1: 1 1 0 0
>>>>> B1: 1 0 1 0
>>>>> A2: 0 0 1 1
>>>>> B2 0 1 0 1
>>>>> A1>B1: 0 1 -1 0
>>>>> A2>B2: 0 -1 1 0
>>>>> The conjunction of these two is 0 since they can't overlap!!!
>>>>>
>>>>> A1>A2: 1 1 -1 -1
>>>>> B1>B2: 1 -1 1 -1
>>>>> The conjunction of these two is possible since they can overlap.
>>>>>
>>>>>
>>>>> Best Regards, Donald McLaren
>>>>> =================
>>>>> D.G. McLaren, Ph.D.
>>>>> Postdoctoral Research Fellow, GRECC, Bedford VA
>>>>> Research Fellow, Department of Neurology, Massachusetts General Hospital and
>>>>> Harvard Medical School
>>>>> 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 Tue, Sep 13, 2011 at 5:07 AM, Cyril Pernet <[log in to unmask]> wrote:
>>>>>> Hi Bob
>>>>>>
>>>>>> looks ok - once you 'run' make sure the subjects factors appear in the
>>>>>> design matrix (as the factor_matrix.png only showed the factors)
>>>>>> conjunctions will be between contrasts 1 0 -1 0 & 0 1 0 -1 and between 1 -1
>>>>>> & 1 -1
>>>>>>
>>>>>> Hope this helps
>>>>>> Cyril
>>>>>>
>>>>>>
>>>>>>> Dear SPM experts,
>>>>>>>
>>>>>>> I am trying to setup a group-level model to test the conjunction of
>>>>>>> simple effects. Namely, I have a 2 X 2 within-subjects factorial
>>>>>>> design, with factors A (two levels, A1 and A2) and B (two levels, B1
>>>>>>> and B2). My current understanding is that it is most appropriate to
>>>>>>> model this using a flexible factorial design, with the following
>>>>>>> factors:
>>>>>>>
>>>>>>> 1. Subject
>>>>>>> 2. Factor A
>>>>>>> 3. Factor B
>>>>>>>
>>>>>>> I've used the attached parameters (pictured in params.png) and factor
>>>>>>> matrix (pictured in factor_matrix.png). As you can see, I've defined
>>>>>>> only the interaction among A and B. In this model, my primary interest
>>>>>>> is in interrogating the conjunction of the simple effects, e.g. A1>
>>>>>>> B1& A2> B2, and also A1> A2& B1> B2.
>>>>>>>
>>>>>>> Is this model correct? I realize this may be a simple-minded question,
>>>>>>> but I'm still living in the days of the one-sample t-test and am
>>>>>>> unsure about how to appropriately use the flexible factorial model.
>>>>>>>
>>>>>>> Sincerely,
>>>>>>>
>>>>>>> Bob Spunt
>>>>>>> Postdoctoral Fellow
>>>>>>> Social Cognitive Neuroscience Lab - www.scn.ucla.edu
>>>>>>> Department of Psychology
>>>>>>> University of California, Los Angeles
>>>>>>
>>>>>>
>>>>>> --
>>>>>> The University of Edinburgh is a charitable body, registered in
>>>>>> Scotland, with registration number SC005336.
>>>>>>
>>>>>
>>>>
>>>
>>
>
|