Hi David
You create F contrast and thus ess files -- these are sum of squares and thus have no positve/negative values. Second level must be done on con files, combinations of betas which have real +/- values.
Having said that, you cannot do a t stat to combine the different cosines (summing would be silly), and indeed using a F test using the identify matrix makes sense as you want to see the contribution of any of those basis function. The only thing I can think of is to set up an ANOVA model with all the betas (57 of them by the look of your F contrast) for all subjects ; the design would have 58 columns (adding the constant) and then a F test here should tell you overall where does it fit. You can then check individual contributions which relate to frequencies.
Cyril
N = 200; % number of scans
TR = 2; % TR in {s}
h = [0.01 0.08]; % {Hz}
n = fix(2*(N*TR)*h + 1);
X = spm_dctmtx(N);
X = X(:,n(1):n(2));
save(fullfile(dataFolder,'DCT.txt'),'X','-ascii');
matlabbatch{1}.spm.stats.fmri_spec.dir = cellstr(glmFol);
matlabbatch{1}.spm.stats.fmri_spec.timing.units = 'scans';
matlabbatch{1}.spm.stats.fmri_spec.timing.RT = TR;
matlabbatch{1}.spm.stats.fmri_spec.sess.scans = cellstr(spm_select('ExtFPList' ,dataFolder, '^sw.*\.nii$',Inf));
matlabbatch{1}.spm.stats.fmri_spec.sess.multi_reg = cellstr(fullfile(dataFolder,' DCT.txt')); %Include DCT basis functions
matlabbatch{1}.spm.stats.fmri_spec.mthresh = 0;
matlabbatch{1}.spm.stats.fmri_spec.mask = maskFile;
matlabbatch{1}.spm.stats.fmri_spec.sess.hpf = 200;
% SPM estimation
matlabbatch{2}.spm.stats.fmri_est.spmmat = cellstr(fullfile(glmFol,'SPM. mat'));
%Create F-contrast
matlabbatch{3}.spm.stats.con.spmmat = cellstr(fullfile(glmFol,'SPM. mat'));
matlabbatch{3}.spm.stats.con.consess{1}.fcon.name = 'cosine';
matlabbatch{3}.spm.stats.con.consess{1}.fcon.weights = eye(57);
matlabbatch{3}.spm.stats.con.consess{1}.fcon.sessrep = 'none';
matlabbatch{3}.spm.stats.con.delete = 1;