SnPM Documentation:

The home of SnPM.

A presentation I gave on SnPM. (4/23/02)

A (much better) presentation by Tom Nichols, one of creators of SnPM.

A chaper of Andrew Holmes' 1994 thesis pertaining to non-parametric statistics and brain imaging.

SnPM Vocabulary:

Exchangeability - The ability to exchange the labels/ordering under the null hypothesis.

Randomization - Randomized conditions balanced across subjects which are interchanged to find adjusted t-values.

Permutation - Computationally the same as randomization, but fail to account for order effects.

Contact me if you would like me to add any more, asfox@wisc.edu

SnPM Capabilities:

MSA2xPerm - Multi-Subject two condition no replication design
Assumes symmetrical distributions for each condition.

MG2i - two group, two condition interaction analyses where there is just *one* scan per condition
Assumes subject (under H0) and scan exchangeability

MG2x - two group analyses where there is just *one* scan per subject
Assumes between subjects anatomical equality.

MS1 - one group analyses where there is just *one* scan per subject (one sample t-test)
Assume under H0 that all images are exchangeable, mean 0, and each voxel has a symmetric distribution.

MSA2x - multi-subject, two condition with replication design (balanced ordering of conditions) where the condition labels are permuted over subject

SSA2x - single subject, two condition activation (with replication)

SSC - single-subject, correlation design

MSC - multi-subject, correlation (Single subject is supported as well)