Hi Rosalia,

If this is not FEAT, then you may consider some changes. The leftmost column (the one full of "1" and "2") should contain just "1" for all subjects, because for randomise, this represents exchangeability blocks (EBs). Also, there's in general no need to split nuisance variables according to experimental group, so age, sex, school and HTA can each be coded into its own single EV.

Demeaning these nuisance won't make any difference as the intercept is already modelled across EV1 and EV2 (which you need to leave unchanged for these contrasts).

All the best,

Anderson

2015-01-25 17:56 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:

Hi Anderson,

Is there any problem is I have demeaned the variables....I am just adjusting for them...so I think it would not have been any problem...right?

I am not working with fMRI data, just with DTI data and running randomise....

Thank you,

Rosalia.

2015-01-25 18:40 GMT+01:00 Anderson M. Winkler <[log in to unmask]>:

Hi Rosalia,

There should be no demeaning in this design, and it can be run in FEAT as is.

All the best,
Anderson

2015-01-25 17:31 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:

Dear Anderson,

Here it is the design.

Kind regards,

Rosalia.

2015-01-25 17:35 GMT+01:00 Anderson M. Winkler <[log in to unmask]>:

Hi Rosalia,

It depends. Sometimes we want to demean EVs, sometimes we don't. It also depends on how the groups were coded and other factors. Maybe if you show the design or give more details it'd be simpler to comment.

Thanks

All the best,

Anderson

2015-01-24 17:54 GMT+00:00 Rosalia Dacosta Aguayo <[log in to unmask]>:

Dear FSL experts,

I have a question regarding covariating for dichotomic variables (1 or 0) when running randomise.

Should I sum all the 1 in the group A and make average of them and substract the average to every 1?

Thank you,

Rosalia.