Dear FSL experts,
I'm currently trying to replicate the findings with FMRILM (of
FMRIstat package) using FEAT. The command in question is:
[df, spatialav] = fmrilm(infname, [repmat('stats/', ncontrasts, 1) ...
strvcat(contrastnames)], xcache, contrasts, exclude,
whichstats, fwhmcor, ntrends, confounds)
in fmrilm.m, usage is shown as
% [DF, SPATIAL_AV] =
% FMRILM( INPUT_FILE, OUTPUT_FILE_BASE, X_CACHE, CONTRAST
% [, EXCLUDE [, WHICH_STATS [, FWHM_COR [, N_TRENDS [, CONFOUNDS
% [, CONTRAST_IS_DELAY [,NUM_HRF_BASES [, BASIS_TYPE [,NUMLAGS
% [, DF_LIMIT ]]]]]]]]]] )
% NUM_HRF_BASES is a row vector indicating the number of basis functions
% for the hrf for each response, either 1 or 2 at the moment. At least
% one basis functions is needed to estimate the magnitude, but two basis
% functions are needed to estimate the delay. If empty (default), then
% NUM_HRF_BASES is 2 for each response where CONTRAST is non-zero and
% _del appears in WHICH_STATS, otherwise NUM_HRF_BASES = 1. By setting
% NUM_HRF_BASES = 2 you can allow for an unknown delay, without
% actually estimating it. Example:
% CONTRAST=[1 -1 0 0; 1 0 -1 0]; WHICH_STATS=['_mag_t'; '_del_t']
% The first contrast is the magnitude of response_1 - response_2;
% the second contrast is the delay of response_1 - response_3.
% The default setting of NUM_HRF_BASES is [2 1 2 1]. By setting
% NUM_HRF_BASES=[2 2 2 1] unknown delays are allowed for the second
% response but not actually estimated.
%
% BASIS_TYPE selects the basis functions for the hrf used for delay
% estimation, or whenever NUM_HRF_BASES = 2. These are convolved with
% the stimulus to give the responses in Dim 3 of X_CACHE.X:
% 'taylor' - use hrf and its first derivative (components 1 and 2), or
% 'spectral' - use first two spectral bases (components 3 and 4 of Dim 3).
% Ignored if NUM_HRF_BASES = 1, in which case it always uses component 1,
% i.e. the hrf is convolved with the stimulus. Default is 'spectral'.
What i'm particularly confused is the num_hrf_bases and basis_type.
Since my command didn't specify the basis_type, the basis_type is
'spectral' using FMRIstat package. Which convolution in FEAT is this
equivalent to? Thank you for your help.
Yoon Han
|