Dear Caroline,
In your first email, you mentioned being interested in testing for the
group by behavioural skill interaction - this would be a contrast [0 0 1
-1].
For the first two contrasts you mention below, you could simply use a
one-sample t-test with covariate, modelling a single group at a time,
with contrast [0 1]. For a two-sample t-test with covariates, the
contrasts would be [0 0 1 0] and [0 0 0 1].
For the main effect of group, that's [1 -1] or [1 -1 0 0] but see the
problem of interpretation in the presence of a significant interaction
in Jeanette's webpage.
Best regards,
Guillaume.
On 30/08/17 15:07, Caroline Landelle wrote:
> Dear Guillaume,
> Thank you for your reply.
> Just to be sure, are these contrats right for the two sample t test including a covariable ?
> * Correlation G1 and covariate [1 0 1 0] (ie [G1 G2 interaction1 interaction2])
> * Correlation G2 an covariate [0 1 0 1]
> * G1>G2 [0 0 1 -1]
>
>
> Best regards
> Caroline
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|