You can simply find all the NaNs and change them to 0. idx = find(isnan(data)); data(idx) = 0; This might be helpful. JR On 2005/5/24, at 上午 11:25, Daniel Kelly (AKA Jack) wrote: > Dear SPM community, > > Whilst trying to develop a new script to specify my model in SPM2, > I keep running into this error message (the line at the top is the > line in my script which causes the problem): > > >> SPM.xCon(c+n_defCntr) = spm_FcUtil('Set',c_name > {c},'T','c',c_wgt{c},SPM.xX.xKXs) > Output from spm_FcUtil.m line 879: > > c = > > 0 > 0 > 0 > 0 > 0 > 0 > 0 > 0 > 0 > 1 > 0 > > ------------ > Output from spm_FcUtil.m line 880: > > c = > > 0 > 0 > 0 > 0 > 0 > 0 > 0 > 0 > NaN > 1 > 0 > > ------------ > ??? Error using ==> svd > NaN or Inf prevents convergence. > > Error in ==> rank at 15 > s = svd(A); > > Error in ==> spm_FcUtil>sf_is_T at 886 > if rank(c) > 1 | any(any(c'*c < 0)), boul = 0; end; > > Error in ==> spm_FcUtil at 210 > if Fc.STAT=='T' & ~sf_is_T(sX,c) > > Weirdly, this problem is intermittent. If I run the command at the > top of this code-dump over and over again, I get this error message > about half the time, even though all the inputs are the same. > > I've been using this batch script for several months without any > problems. It's only recently that it's started to play up. The > only difference between my error-prone batch script and my previous > batch scripts is the number of regressors and contrasts. This new, > error-prone batch script has 10 regressors. > > Anyone got any clever ideas? Please? All advice will be very > greatfully received. > > I'm using SPM_devel from about a month ago, running on Matlab 7SP2 > on Gentoo Linux. My MEX files are re-compiled using ICC for my P4. > > Many thanks, > Jack > ===================================== > Jack Kelly > nickname = Jack > birthname = Daniel > > PhD Student > Institute of Cognitive Neuroscience > University College London > > www.XLK.org.uk > =====================================