Thank you for your answer Donald.

I am a little bit confused about how to introduce the factor "Subject". The levels of this factor would be the number of subjects that I have at "Pre" or the total number of acquisitions per group? I have difficulties to figure it out.

In my 3-factor's design, in the 1st level analysis for each acquisition (pre and post) of each subject, I made the following subtraction :

S1=Sa-Sbaseline and S2=Sb-Sbaseline.

Then I introduced the "con" images of S1 and S2 of each subject and acquisition to each cell (PreG1S1 (6 cons), PreG2S1 (6 cons) ...) until 24 (6 cons per cell). With one more factor, would I have 1 con per cell i.e. PreG1S1Subj1 (1con), PreG1S1Subj2 (1 con) ...?

Thank you for your help

Yolanda

2013/11/21 MCLAREN, Donald <[log in to unmask]>

On Thu, Nov 21, 2013 at 8:20 AM, Yolanda Vives <[log in to unmask]> wrote:

Dear all,

I have not a lot of experience with factorial designs, sorry if my questions are evident. I have an fMRI 2nd level 2x2x2 full factorial design, where

Factor 1 (longitudinal) : 2 levels (Pre and Post)

Factor 2 (group) : 2 levels (G1 and G2)

Factor 3 (stimuli) : 2 levels (S1 and S2)

You need to use the flexible factorial model and you need to include subject, its been awhile since I tried having 4 factors (Factor 4 subject), so I am not sure if it is possible. You might need to use GLM Flex.

An alternative would be to remove 1 of your factors by forming the difference of the images before creating the model.

 

I have two questions:

1.- I understand that the levels of my factor 1 (pre and post) are dependent taken into account that I have 2 measurements of the same subjects. Factor 3 corresponds to 2 different stimuli (S1 and S2), are these levels also dependent or would they be independent?

If the observations are from the same subject, then they are dependent. Thus, if you measure S1 and S2 in the same subjects, then they are dependent.

 

2.- I have some difficulties setting the contrasts. I have the following comparisons, how would I define the contrasts to compare X>Xpost? Is it right to use a full factorial design or should I do it in a different way?

You need to use the flexible factorial or GLM Flex. You should put the main effects or the factors into the model along with subject.

In the example below, X>Xpost would be 1 0 0 -1 0  -1 0 1 0 because you subtract Xpost from the X contrast. This model and contrast should not be used though as you need to use the flexible factorial or GLM Flex.

 

 

	PreG1S1	PreG2S1	PostG1S1	PostG2S1	PreG1S2	PreG2S2	PostG1S2	PostG2S2
G1Pre (S1 > S2) = X	1	0	0	0	-1	0	0	0
G1Post (S1 > S2) = Xpost	0	0	1	0	0	0	-1	0
X > Xpost

Thank you for your help,

Yolanda

--
Yolanda Vives
PIC (Port d'Informació Científica)
Campus UAB, Edificio D
E-08193 Bellaterra, Barcelona
Telf. +34 93 586 8232
http://www.pic.es
Avis - Aviso - Legal Notice: http://www.ifae.es/legal.html