Dear Dudley, In this instance, you should set 'Variance' to be 'Equal' in your two-sample T-test. Best regards, Guillaume. Dudley Ibbett wrote: > Hi > > I am trying to set up a simple SPM unpaired T test comparing a single > patients TcHMPAO SPECT data set with a 20 controls (studies that have > been reported as "normal"). I am using SPM8b with the following basic > model parameters: Two Sample T-test, Threshold Masking: Relative, Global > Calculation: Mean, Global Normalistation: Yes, Normalisation: Proportional. > > I have found an article on the web by Rik Henson (MRC) on this subject > and the Two-Sample T-test appears to be recommended. This articles > suggests I should be able to use the T-contrast [1 -1] to test the > control group against the single patient, however SPM tells me this is > and invalid contrast. The only contrast it allows is [0 1]. I would be > very grateful if someone could explain what I am doing wrong. > Presumably I should be able to set up a contrast that will test > for areas where the single patient is greater that the control group and > also a contrast that test for areas where the single patient is less > than the control group. > > Many Thanks > > Dudley -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG