Hi Donald,
Thank you for your answer regarding my mixed model design. I´ve been
reading the tutorial by Glascher and Gitelman (2008) but still have doubts
about how to specify the diferenttial contrasts in the "first level" of VBM
so that I will be able to test for the effect of the interactions that I
mentioned in the flexible factorial. Could you please provide me with a
short description on how to proceed or refer me to the pertinent
literature, please?
This would be extremely helpful.
Thank you in advance for your time.
Best,
Noelia
2018-06-11 16:43 GMT+03:00 MCLAREN, Donald <[log in to unmask]>:
> HI Noelia,
>
> Those are all fine as they are group*time interactions.
>
> It's only the effect of group that isn't valid.
>
> Best Regards,
> Donald McLaren, PhD
>
>
> On Mon, Jun 11, 2018 at 6:06 AM, Noelia Martinez-Molina <
> [log in to unmask]> wrote:
>
>> Hi Donald,
>>
>> Thank you for clarifying that in the pooled model SPM only provides the
>> within-subject error term. I was then hoping to set up a partitioned error
>> model using a flexible factorial design. As a reminder, I´m running a
>> longitudinal VBM analysis with 2 groups (A, B) and 3 time points (1,2,3)
>> and would like to test for the following interactions:
>>
>> (A>B)x(1>2)
>> (A>B)x(2>3)
>> (A>B)x(1>3)
>>
>> However, in VBM there is no first level and I would appreciate if you
>> could provide some help about how to set the differential contrasts and
>> then the flexible factorial (including the contrast matrix) to test for the
>> effect of the aforementioned interactions.
>>
>> All the best,
>> Noelia
>>
>>
>>
>> 2018-06-08 17:18 GMT+03:00 MCLAREN, Donald <[log in to unmask]>:
>>
>>> Hi Noelia,
>>>
>>> The main effect of group is an invalid contrast in a repeated measures
>>> ANOVA using a GLM, even if you didn't get the error message. The reason is
>>> the error term for this model is the within-subject error term not the
>>> between-subject error.
>>>
>>> For the main effect of group, you want to use a two sample t-test of the
>>> average of the two visits.
>>>
>>> Also, covariates are not valid in a repeated-measures ANOVA. If you want
>>> to use covariates, you need a mixed effects model.
>>>
>>>
>>> Best Regards,
>>> Donald McLaren, PhD
>>>
>>>
>>> On Fri, Jun 8, 2018 at 8:21 AM, Noelia Martinez Molina <
>>> [log in to unmask]> wrote:
>>>
>>>> Dear SPM experts,
>>>>
>>>> I already run a flexible factorial design with 2 groups (n1=17, n2=12)
>>>> and 2 within-subject conditions. This is a VBM longitudinal analysis and
>>>> here are the factors:
>>>> Factor 1: Subject
>>>> Factor 2: Group (2 Levels)
>>>> Factor 3: Time (2 Levels)
>>>> Covariate1: age
>>>> Covariate2: total intracranial volume
>>>>
>>>> I set up the following contrasts:
>>>>
>>>> F tests:
>>>>
>>>> Time differences in Group1: 1 -1 0 0
>>>> Time differences in Group2: 0 0 1 -1
>>>> Main Effect Time: 1 -1 1 -1
>>>> Main Effect Group: ones(1,2)/2 -ones(1,2)/2 0 0
>>>> ones(1,17)/17-ones(1,12)/12
>>>> Interaction Time by Group: 1 -1 -1 1
>>>>
>>>> I think all of these should be correct but when I entered the contrasts
>>>> weights for Main Effect of Group, I got the "Invalid contrast" error.
>>>>
>>>> Could anyone explain me why this is happening and what would be the
>>>> correct contrast weights matrix for this, please?
>>>>
>>>> Thank you in advance for your help.
>>>>
>>>> Best,
>>>> Noelia
>>>>
>>>
>>>
>>
>>
>> --
>>
>> *****************************************************
>> Noelia Martínez Molina, PhD
>> Postdoctoral Researcher
>> Cognitive Brain Research Unit (CBRU)
>> Department of Psychology and Logopedics
>> Faculty of Medicine
>> University of Helsinki
>> E-mail: [log in to unmask] <[log in to unmask]>
>> Phone1: +358 50 31 88 048 (Finnish)
>> Phone2: +34 620 15 23 87 (Spanish)
>>
>> -------------------------------------------------------------
>> *"Where words fail, music speaks".*
>> H.C. Andersen
>> ******************************************************
>>
>
>
--
*****************************************************
Noelia Martínez Molina, PhD
Postdoctoral Researcher
Cognitive Brain Research Unit (CBRU)
Department of Psychology and Logopedics
Faculty of Medicine
University of Helsinki
E-mail: [log in to unmask] <[log in to unmask]>
Phone1: +358 50 31 88 048 (Finnish)
Phone2: +34 620 15 23 87 (Spanish)
-------------------------------------------------------------
*"Where words fail, music speaks".*
H.C. Andersen
******************************************************
