Dear Feng,
The default for a two sample t-test in the SPM interface is to assume
unequal variance so, to get the same result, you should change it to
equal. There is also a small adjustment for the variance estimate (see
towards the end of spm_spm.m, ResMS modification) so make sure the
estimation takes place with the fMRI defaults where this adjustment is
disabled.
Best regards,
Guillaume.
On 14/05/18 02:55, chenhf_uestc wrote:
> Dear SPM experts,
>
> I perform two-sample t tests in SPM12 with two groups of images without
> any covariates (1. specify 2nd level: select two-sample t-test, and
> input the two groups of images; 2. estimate the mode; 3. input the
> contrast [1 -1]) and I obtained the t map. Next, I used MATLAB function
> "ttest2" to perform two-sample t tests voxel-by-voxel, and the t map
> could also be obtained. However, the t values in these two t maps are
> not exactly equal.
>
> Subsequently, I try to use DPABI software (http://www.rfmri.org/dpabi)
> to perform two-sample t tests, and actually, it also based on regression
> model like SPM (attached file). The t map obtained by DPABI is the same
> as the map obtained by MATLAB function "ttest2".
>
> I am wondering why SPM's t map is slightly different from MATLAB
> function "ttest2" and DPABI software.
>
> Any help would be greatly appreciated!
>
> Best,
> Feng
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|