Print

Print
Hi Sandra,

Depends. If baseline was at zero, the interaction tells where there is a response when both A and B happen simultaneously, but not where either alone happen. For the full 4 cells, need to center the EV, and to find the direction you'd look at the signs of the main effects (which go up, which go down, even if not significant). There are other threads in the mailing list in which this is discussed at higher levels. The same principle applies to 1st level.

All the best,

Anderson


On 28 September 2016 at 14:31, Thijssen, S. <[log in to unmask]> wrote:

Thank you, Anderson! If I find a significant interaction effect, how do I know which cells are significantly different? (e.g. is there an opposite effect of B for low and high levels of A, or does B have an effect only on low(or high) levels of A).

 

Thanks,

Sandra

 

From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Anderson M. Winkler
Sent: zondag 25 september 2016 10:27
To: [log in to unmask]
Subject: Re: [FSL] 2 x 2 first level analysis

 

Hi Sandra,

 

Mark's advice from the first thread still holds. If you wanted the 4 cells (i.e., interaction also considers both baselines as in the original thread), you could "center" the EVs when creating the interaction; there is an option for that, or you could do it manually before entering the values.

 

All the best,

 

Anderson

 

 

On 23 September 2016 at 14:41, Thijssen, S. <[log in to unmask]> wrote:

Dear experts,

 

I was wondering if you could help me out with the design of my analysis.  I am conducting an fMRI task which examines 2 factors (say A and B). Each factor has 2 values (A1 and A2, and B1 and B2), which results in 4 conditions: A1B1, A2B1, A1B2, and A2B2. We are interested in both main effects and in the interaction effect of A and B.

 

Based on previous threads, I came up with two options on how to analyze the data and I was hoping you could let me know which option is best.

 

A.      I can make an EV for a main effect of A and a main effect of B and from these EVs create and interaction EV. I will create 4 contrasts, 1 for each main effect and a positive and negative contrast for the interaction EV (based on: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind04&L=fsl&P=R11865&1=fsl&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4 )

B.      If I correctly understand the information in https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind05&L=fsl&P=R39478&1=fsl&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4 I could also make an EV for each condition (so A1B1, A2B1, A1B2, and A2B2) and test the interaction effect by multiplying the relevant EVs in a contrast 

 

Thank you for your help!

 

Best,

Sandra