Hi - not sure what you mean by "considered"?   Permutation testing generates an empirical null distribution for any test statistic and hence parametric modelling assumptions (e.g. influence of DoF) are bypassed and automatically accounted for.

Cheers.




On 4 Nov 2012, at 15:22, Andi Heckel wrote:

Dear List,

randomise offers the -D option, i.e. it demeans the input data before model fitting.
Alternatively, one has to model the mean as a regressor.

As I understand, this regressor - as any linearly independent regressor - would cost 1 degree of freedom.
I was wondering, whether this 1 degree of freedom is considered by randomise when omitting the mean regressor and using the -D option instead. (could not find the answer in the docs).

THX + kind regards,
andi


---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------