Term	Definition
Action unit	Smallest unit of visible facial movement, based on underlying muscle distribution
Unconditional probability	Proportion of events in which a unit or combination of units occurs in a condition or dataset: $P\left(A\right)=\frac{NumberofoccurrencesofA}{NumberofpossibleoccurrencesofA}$
Conditional probability	Proportion of events in which a unit occurs, given that another unit is already present: $P\left(B|A\right)=\frac{NumberofoccurrencesofAandBtogether}{NumberofoccurrencesofA}$
Specificity	Proportion of events in which a condition is observed if an element is present; strength of evidence that AU is connected to a condition rather than another: $P\left(C1|A\right)=\frac{NumberofoccurrencesofAinContext1}{NumberoftotaloccurrencesofA}$
Bootstrap	A resampling method in which cases of a condition are repeatedly randomly selected with replacement and the statistic of interest is calculated for each iteration, to provide estimates of the distribution of the statistic in the sample
Permutation	A resampling method in which the null distribution of a statistic of interest is created by randomly shuffling some aspect of the original distribution repeatedly
Node	Element of the network, in this case the AUs
Edge	Connection between elements/nodes in a network; e.g., based on the co-occurrence of AUs or on their conditional probability
Weighted network	Network in which the edges between nodes can take different values depending on how weakly or strongly the nodes are connected
Unweighted network	Network in which the edges between nodes are either 1 (present) or 0 (absent)
Directed network	Network in which the connections between two nodes can be asymmetrical, i.e., A → B ≠ B → A
Undirected network	Network in which the connections between two nodes is symmetrical, i.e., A ↔ B
Bipartite graph	Network in which nodes from two different categories (e.g., condition and AU) can be connected between categories (AU→ Condition), but not within categories
Node cluster	Structural element of a network, group of nodes that have strong connection with each other, but weak connections with other nodes outside the cluster
Transitivity	Triads of nodes are considered transitive if all three nodes are connected with each other; a network has high transitivity if a large number of triads are closed ( A → B → C → A) compared to triads that just show two connections (e.g. A → B, B → C)
Density	Proportion of the potential connections between nodes that are actual existing connections. In a dense network, all or most connections are present; in a sparse network, only few connections are present
Degree	Number of actual connections a node has with other nodes in a network. Nodes with high degree co-occur frequently with other nodes
Strength	Mean weight of connections of a node with all other nodes in a weighted network