On small sample experiments in neuroimaging

Most human brain imaging experiments involve a number of subjects that is unusually low by accepted statistical standards. Although there are anumber of practical reasons for using small samples in neuroimaging we need to face the question regarding whether results obtained with only a fewsubjects will generalise to a larger population. In this contribution we address this issue using a Bayesian framework, derive confidence intervals forsmall samples experiments, and discuss the issue of the prior.