 |
 |
Hi, The order number determines how flexible the shape is within the respiratory period. If you only use order=1 then you just get a basic sinusoid shape, although in the case of the respiratory EVs they are also modulated by the depth of the breathing in that period to some extent. However, higher orders give the method more flexibility to fit different shapes to the measured respiratory changes. There is no way to determine in advance what the "right" number is. Using higher orders increases the number of EVs (and so you lose some DOF in the stats, but that is usually negligible when you start with hundreds of timepoints). There are diminishing returns for adding more and more EVs (i.e. increasing the order) and you can get a feeling for this in your data by using a higher order and then looking at how much variance each order explains (this can be done by entering the EVs manually as main EVs, not via the confounds, and then putting F-tests over the different orders). If the F-test is significant at a voxel then it says that this order is explaining more variance that you'd expect by chance. Once you see an order where the F-test is rarely significant (as each voxel is tested separately, so some will always be significant on a voxel-level given the false positive rates, so take this into account) then you can conclude that using this order was not really useful. I hope this helps. All the best, Mark On 27 Jun 2014, at 01:11, Tae-Ho Lee <[log in to unmask]> wrote: > I found the answer about this question from the manual. > I should have read the guide more carefully. Please ignore the previous post! > > But another question is that how I can decide the order number (I'm sorry to keep bothering you all). > I guess that it would be enough to make respiration only regressors with the default order value (i.e., 1), right? > > Thanks for any feeds! > > Best, > TH