> After running the 2-nd level analysis (one sample t-test) I get this
error
> when I select VOLUME to have the statistic table; in the CLUSTER-LEVEL
P
> CORRECTED coloumn I see NaN for three times.
>
> Warning: Returning NaN for out of range arguments
> > In c:\spm2\spm_Pcdf.m at line 84
> In c:\spm2\spm_P_RF.m at line 102
> In c:\spm2\spm_P.m at line 44
> In c:\spm2\spm_list.m at line 453
> Warning: Returning NaN for out of range arguments
> > In c:\spm2\spm_Pcdf.m at line 84
> In c:\spm2\spm_P_RF.m at line 102
> In c:\spm2\spm_P.m at line 44
> In c:\spm2\spm_list.m at line 453
> Warning: Returning NaN for out of range arguments
> > In c:\spm2\spm_Pcdf.m at line 84
> In c:\spm2\spm_P_RF.m at line 102
> In c:\spm2\spm_P.m at line 44
> In c:\spm2\spm_list.m at line 453
>
> What doest it mean? Is my analysis corrected? How shall I consider the
NaN
> clusters?
I saw your results, and I noticed you only had df=2. Probably spm_P_RF.m
function does not work properly under such low df.
On top of that, FWHM was [0.9x0.9x0.9] voxels, and that makes clusters
very small in general. So that may be another reason for NaN. I presume
you didn't do any smoothing, correct?
Besides the validity of your p-values, I think you should consider
including more subjects (for larger df) and smoothing the data (for
larger FWHM). Otherwise, the results do not seem very convincing.
-Satoru
|