Hello,
I had some questions regarding doing basic statistics using the general
linear model implemented in SPM. Basically, I am trying to compare
(voxel by voxel) single patients to a group of controls (10) using a
two sample t-test using the Basic Models option of SPM99. My results
seem to be correct (based on SPM99's correct identification of lesion
locations) but I am not quite sure I understand exactly what is being
computed. The standard formula for the T statistic for a two sample
t-test is given by
(mu1-mu2) / sqrt(s1^2/n1 + s2^2/n2)
where the mu's are the sample means, the s's are the sample variances,
and the n's are the sample sizes.
How can the computation work with a sample size of 1 for one of the
samples as this would result in a divide by zero when computing the
sample variance. Does the SPM99 model assume that the sample variances
are equal? If so, is the variance being computed by pooling all 11
subjects' data (10 controls + 1 patient)?
Thanks in advance for any advice on what is probably a simple question
for most of you.
Sunil
|