 |
 |
Hi, On 10 Dec 2008, at 18:03, Shih-Wei Wu wrote: > Thanks a lot. > Just to further clarify: > 1. In the case of seeing if there is an effect of each parametric > regressor, and say that I have 3 parametric regressors and set up > the contrasts -- > contrast 1: (1 0 0) > contrast 2: (0 1 0) > contrast 3: (0 0 1) > > rescaling or normalizing the value of each parametric regressor > should not matter here because z is the normalized score of the > parameter estimate. > > But rescaling/normalization is important when I want to compare > different regressors. Say that I want to compare the parameter > estimate of regressor 1 and 2 and therefore set up a contrast (1 -1 > 0). > > Is that right? That's all correct - scaling isn't an issue in single-EV contrasts, but can be for more-than-one-EV contrasts. > 2. When you say that the EVs should be of height 1, do you mean that > the maximum height should be of 1? Should I rescale them so that > they all have the range [-1,1]? Normally the raw EV waveforms (pre-convolution, e.g. as input via either custom timings format), whether event-related or block, go from 0 (baseline) to 1 (activation) and so have a range 0:1. Cheers. > > thanks > shihwei > > On Wed, Dec 10, 2008 at 4:10 AM, Steve Smith <[log in to unmask]> > wrote: > Hi - yes this is important otherwise you can't easily interpret > contrasts between conditions. The normal way is just to make sure > that the EVs are of height 1 as input to FEAT, i.e. before > convolution, etc., and then this is all comparable across EVs. > > Cheers. > > > > On 9 Dec 2008, at 01:42, shih-Wei Wu wrote: > > Dear all, > > A question with regard to scaling of parametric regressors. > > I have 3 parametric regressors that occupy different time points > during a trial. The range of > values between the regressors differ. One regressor has range > approx. [-1,1] with mean at > 0, the second regressor has range [-3,3], and the third regressor > [-5,5]. > > What would be the correct way to re-scale them so I don't have to > worry about scaling > problems when running the GLM? > > thanks! > shihwei > > > > --------------------------------------------------------------------------- > 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 > --------------------------------------------------------------------------- > --------------------------------------------------------------------------- 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 ---------------------------------------------------------------------------