Print

Print
Hi Carole,

From earlier emails it seems you have 2 groups. Do you mean that you have 2 sets of 3 scans for each monkey, being 3 scans without treatment and the other 3 with treatment, and this was repeated for an overall sample of 3 monkeys?

If yes, the only analysis I can think of that won't be too terribly hampered by the small sample size is this:
1) Run melodic with temporal concatenation using all 6x3=18 scans.
2) Run the dual regression.
3) Run randomise using design.mat, design.con and design.grp as follows:

design.mat: should contain 18 rows and 4 columns:
- EV1: Code +1 for treatment and -1 for placebo
- EV2: Code 1 for monkey 1, 0 for others
- EV3: Code 1 for monkey 2, 0 for others
- EV4: Code 1 for monkey 3, 0 for others

design:con:
- C1: [1 0 0 0] (this tests treatment > placebo)
- C2: [-1 0 0 0] (this tests treatment < placebo).

design.grp:
This is a single column that defines one exchangeability block per monkey:
1
1
1
1
1
1
2
2
2
2
2
2
3
3
3
3
3
3

Then when running randomise, make sure that the options "-d design.mat -t design.con -e design.grp" are all included. There should be 8000 permutations possible.

Regarding head motion, please see this other thread: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;448bba7e.1502
I wouldn't include it.

All the best,

Anderson



On 16 March 2015 at 07:47, carole Guedj <[log in to unmask]> wrote:

__________________________

Carole Guedj, PhD Student

Impact (Integrative multisensory perception Action cognition team)
Centre de Recherches en Neurosciences de Lyon - CRNL
INSERM U1028 - CNRS UMR5292 - Lyon 1 University
16, avenue du Doyen Lépine
69500 Bron, France

Phone: +33 (0)4 72 91 34 35


Dear Anderson,

Sorry, let me be more precise. I tested the effect of one drug in 3 monkeys, at rest. For several reasons link to monkey research it is not possible for me to have more than 3 subjects in my study. 
So I was very interested to use the Dual-regression with Randomise to analyse this data. But as you said, it is not easy to get proper statistical results with only 3 subjects (even if I multiply the number of scan session to get a bigger sample).
It was the reason why I would like to run the dual-regression separately on my 3 subjects (using a matrix design with 2 groups: group1 = monkey 1 with the drug, group2 = monkey 1 with the placebo… and then for each monkey). So of I find a common effect on the 3 subjects, it can comfort me in the effect seen on the group. 
I don’t know if it is more clear for you, please tell me if it’s not!

My second question it about the Explanatory Variable that I put in my matrix design. Should I include head motions? (they are already regressed during my preprocessing, but perhaps that would help to better model the data; on the other hand, this means that I have to increase the number of permutation?)



Thousand of thanks for your precious help!

All the best,

Carole

__________________________

Carole Guedj, PhD Student

Impact (Integrative multisensory perception Action cognition team)
Centre de Recherches en Neurosciences de Lyon - CRNL
INSERM U1028 - CNRS UMR5292 - Lyon 1 University
16, avenue du Doyen Lépine
69500 Bron, France

Phone: +33 (0)4 72 91 34 35