Print

Print
See below.

On Fri, Nov 1, 2013 at 10:31 AM, Heleen Hoogeveen <[log in to unmask]> wrote:
Hi all,

Hereby I would like to receive your feedback on the flexible factorial design I made.

Factor 1 : Subject (Independence : yes  / Variance : equal)
Factor 2 : Age group (Young (N = 39) vs. Old (N = 35)) - between subject (Independence : yes  / Variance : unequal)
Factor 3 : Taste condition (Sweet, Sour, Salty, Bitter, Water) - within subject (Independence : no / Variance : equal)

I'm interested in :
1) 'Main' effect of Taste condition ("Which brain-areas show differences in BOLD in response to different taste conditions?") - within subject
2) Interaction effect of Age x Taste condition ("Which brain-areas show age-related differences in BOLD in response to different taste conditions?") - within and between subject

This is considered a within-subject effect.
 

I'm confused about whether or not to specify Subject and Taste as a main effect (get columns for this in your design matrix) in answering these questions (Age is not specified as a main effect, since group effect may not be analysed this way).

You need to have Subject, Taste, and Age as main effects in the model. This way you get the columns you need in the model. Subject columns will have 0s.

I tried this to answer my questions:

1) 'Main' effect of Taste condition - only specified the interaction Age x Taste (no Subject columns)
F-contrast : 1 -1 0 0 0      1 -1 0 0 0
                   0 1 -1 0 0     0 1 -1 0 0
                   0 0 1 -1 0     0 0 1 -1 0
                   0 0 0 1 -1     0 0 0 1 -1


I'd change these to 1/2 and -1/2 since you have 2 groups.

 
2) Interaction effect of Age x Taste condition - specified the Subject columns as well as the interaction Age x Taste
F-contrast : zeros(1,74)  1 -1 0 0 0      -1 1 0 0 0
                   zeros(1,74) 0 1 -1 0 0      0 -1 1 0 0
                   zeros(1,74) 0 0 1 -1 0      0 0 -1 1 0
                   zeros(1,74) 0 0 0 1 -1      0 0 0 -1 1



This would be correct, assuming that subjects appear first in the design matrix. The same design matrix can be used for both questions.
 
After this, I would like to test simple effects (compared two taste conditions, and the Age X Taste interaction with two tastes). Can I use these two model for testing this as well? Or do you test all these effect in a model specifying a columns for Subject (74), Age (2), Taste (5), and the Age x Taste interactions (10) (= 91 columns)?

You should only have one model. It can be used to test pair-wise Taste effects.
 

Many thanks in advance,

Heleen Hoogeveen