The principles of the statistical F-test, used in EXAFS fitting in order to
compare different models and eventually choose the more likely, are summar
ized. Examples, taken in the fields of inorganic chemistry and biochemical
applications are discussed. F-test is shown to be very useful to check the
statistical limits of a model and to discriminate between physically accept
able models. However, this tool is useless when the choice involves physica
lly unrealistic solutions since it cannot prevent parameter variations out
of their physically significant limits.