Dear Marco,
if your models are nested (and they seem to be here), you can use an
F-test as a model comparison device (the extra sum of squares principle)
- you only need to specify a full model and test for any reduced model
using the appropriate F-contrast.
If your model are not nested, you can use bayesian model comparison, see
these two papers for theory and application:
http://onlinelibrary.wiley.com/doi/10.1002/hbm.20327/full
http://journal.frontiersin.org/article/10.3389/fnhum.2011.00037/full
Best regards,
Guillaume.
On 04/07/16 11:13, Marco Valenti wrote:
> Dear experts
>
> I'm at my first experience with a functional study and I've a basic
> question for you.
>
> I'm creating the design matrix and I'm sill evaluating, for example, the
> use of time or parametric modulation or the use of derivatives of
> canonical HRF as regressor. If I create several models, is there a way
> to compare them and evaluate which model is more suitable in
> experimental data description?
>
>
>
> Best Regards
>
>
>
> Marco Valenti
>
> --
> This email was Malware checked by UTM 9. http://www.sophos.com
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|