MEDx Frequently Asked Questions - Statistics
Q: What do -log10(P(nmax>=k)) and P(nmax>=k) represent within the
MEDx cluster detection results table?
P(nmax>=k) is the probability of obtaining a cluster of size k or greater
just by chance alone. nmax is a random variable representing the maximum
size of a cluster in a randomly generated Z map. Imagine generating a
million spatially smooth Z maps at random. The number of Z maps having
a cluster of size k or larger divided by a million will be approximately
P(nmax>=k). This is the notation used in the original paper by Friston,
Worsley, et al.; for more detail,
see that paper .
The Cluster Detection method first thresholds the Z map at the user-selected
Z value threshold. It then visits each cluster and finds the size k of
each cluster. P(nmax>=k) is then calculated for each cluster; if P(nmax>=k)
for a given cluster is below the selected probability threshold, then
that cluster is deemed to be too large to have occurred by chance alone.
-log(P(nmax>=k)) is the negative log of P(nmax>=k). It can be useful
for comparing P values which have a very large range (from 1e-39 all the
way to 1.0, with many intermediate values possible), especially if they
are displayed as a map. Often log transforms are used to "squash" the
range of data, which is why semilog or log paper is sometimes used. Taking
the negative log of the P-level is analogous to measuring acidity with
pH.
Other FAQ Topics
If you have a question that you would like to see addressed in our list
of Frequently Asked Questions, please contact
customer support.
Back to FAQ Home
|
