288 P. Boba and K. Hamacher
Fig. 2.A network with two coupled feedback loops that eventually implement two
(almost) independent clocks. They can, however, influence each other via the interac-
tion parametrized bykvy. Note the positive feedback depicted by an arrow and the
down-regulating interaction illustrated by an square-like arrow head. We show the
parameters used in our subsequent analysis.
Fig. 3.(a) Sample time series for the four molecular species of the intracellular network
of Fig. 2 .(b)u,vafter binning usingpartition data; note, that the overall scales are
adjusted by this procedure.
we clearly see the difference in the TE favoring an interpretation of a coupling
fromvtoyas the system shows by design wherevandyare coupled viakvy.
Note, however, that under a null hypothesis testing (NHST) thepvalues do not
meet the value of∼5% under the well-established Fisher rule. Thus, we would
in most cases reject the TE values anyway; only for larger values ofkvydo we
obtain significance values conforming to the threshold.
Further example applications and detailed usage instructions are discussed
in the supplemental material to this paper [ 3 ].
3 Availability
The code for package TransferEntropyPTis made available under GPL-
license. The current version can be downloaded from http://www.cbs.
tu-darmstadt.de/TransferEntropy. Installation inRcan be achieved via the com-
mandR CMD INSTALL TransferEntropyPTx.y.z.tar.gzwhere x, y, z are the
(sub)version numbers for the downloaded file.