Graph Theory: Trees, Forests & Efficiency

The Basics of Social Network Analysis

Tree Graphs & Forests

What is a Tree Graph?

  • If a graph is both connected and has no cycles (acyclic), then it is a tree graph.
  • The Cycle Restriction:
    • A cycle is a path that begins and ends with the same node.
    • In a tree graph, you can never start with a node and trace a sequence of distinct edges that take you back to where you started!

Properties of Tree Graphs

  • Tree graphs possess several unique mathematical and structural properties:
    1. The Single Path Property:
      • Every pair of nodes in a tree graph is linked by exactly one unique path.
      • There are no redundant or alternative paths between any two actors.
    2. The One-Cycle Property:
      • Adding a single new edge connecting any pair of non-adjacent nodes always creates exactly one cycle, destroying its tree status.

Tree Graph Metrics: Vertices & Edges

  • In any tree graph, the number of edges (\(m\)) is always exactly equal to the number of nodes (\(n\)) minus one: \[m = n - 1 \qquad \iff \qquad n = m + 1\]
  • Sum of Degrees Identity:
    • Recall that the sum of degrees in any undirected graph is \(2m\).
    • Substituting \(m = n - 1\), the sum of degrees of a tree graph is always: \[\sum_i k_i = 2(n - 1) = 2n - 2\]
    • In a tree graph, the sum of degrees is always exactly twice the number of nodes minus two!

Tree Graph Density

  • Because of the strict \(m = n - 1\) edge constraint, tree graph density simplifies beautifully.
  • Density of a Tree Graph: \[d(T) = \frac{2m}{n(n-1)} = \frac{2(n-1)}{n(n-1)} = \frac{2}{n}\]
  • For our 5-node tree graph on the right (\(n = 5\)):
    • The density is: \[d(T) = \frac{2}{5} = 0.40 \quad (40\%)\]

Tree Graph Average Degree

  • Average degree in a tree graph also simplifies beautifully due to the edge constraint.
  • Average Degree of a Tree Graph: \[\bar{k} = \frac{\sum k_i}{n} = \frac{2m}{n} = \frac{2n - 2}{n}\]
  • For our 5-node tree graph on the right (\(n = 5\)):
    • The average degree is: \[\bar{k} = \frac{2(5) - 2}{5} = \frac{8}{5} = 1.6\]

Trees as Minimally Connected Graphs

  • Trees are minimally connected graphs:
    • They are connected using the smallest possible number of edges needed to connect \(n\) vertices.
  • The Bridges Rule:
    • Removing even a single edge from a tree graph completely disconnects the graph.
    • Thus, every edge of a tree graph is a bridge (a minimum edge cutset of size 1)!
  • Disconnecting a tree graph by removing any edge splits it into separate connected components, which are themselves trees.

Numerical Example: 8-Node Tree

Let’s compute the metrics for the 8-node tree shown on the right:

  • 1. Nodes and Edges:
    • Number of nodes: \(n = 8\)
    • Number of edges: \(m = n - 1 = 7\)
  • 2. Degree of each node:
    • \(k_A = 2, k_B = 3, k_C = 3, k_D = 1, k_E = 2, k_F = 1, k_G = 1, k_H = 1\)
    • Sum of degrees: \(\sum k_i = 14\)
    • Check: \(2(n-1) = 2(7) = 14 \quad \checkmark\)
  • 3. Network Density:
    • \(d(T) = \frac{2}{n} = \frac{2}{8} = 0.25 \quad (25\%)\)
  • 4. Average Degree:
    • \(\bar{k} = \frac{2n - 2}{n} = \frac{14}{8} = 1.75\)

Forests: Visualizing Disconnection

  • Forest:
    • A disconnected graph whose components are all tree graphs.
  • Before (Left):
    • A connected 9-node tree graph.
  • After (Right):
    • Removing edge D-H splits the graph into two separate trees (a forest):
      • A bigger tree of 6 nodes on the left (\(\{A, B, C, D, E, F\}\)).
      • A smaller tree of 3 nodes on the right (\(\{G, H, I\}\)).

Special Types of Trees: Line, Star & Caterpillar Graphs

  • The Path Graph (\(P_n\)):
    • A single straight path connecting \(n\) nodes.
    • Used to model sequential diffusion processes (like the “Telephone Game”).
  • The Star Graph (\(S_n\)):
    • A central hub connected to \(n-1\) leaf nodes.
    • Represents extreme structural centralization.
  • The Caterpillar Graph:
    • A central path (“body”) with pendant leaf nodes (“legs”) sticking out of it.

The Graph Efficiency

  • Concept (Krackhardt, 1994):
    • Measures how economical the connectivity of a graph is.
    • A graph is efficiently connected if it uses the minimum possible number of edges to remain connected.
  • Formula: \[F(G) = 1 - \left[\frac{2(m - n + 1)}{(n-1)(n-2)}\right]\]
    • Where \(m\) is the number of edges and \(n\) is the number of nodes.
  • Beyond Trees (Decreasing Efficiency):
    • As edges are added beyond \(m = n - 1\), the numerator \((m - n + 1) > 0\), causing \(F(G)\) to decrease toward \(0.0\).

Graph Efficiency: A Tree Example

  • Numerical Example (The Tree on the Right):
    • Pictured on the right is a tree of \(n = 5\) nodes and \(m = 4\) edges.
    • Substituting \(n = 5\) and \(m = 4\) into the formula: \[F(G) = 1 - \left[\frac{2(4 - 5 + 1)}{(5-1)(5-2)}\right]\] \[F(G) = 1 - \left[\frac{2(0)}{4 \times 3}\right] = 1 - 0 = 1.0\]

Detailed Example: Graph 1 (Less Efficient)

  • Graph 1 Structure:
    • Contains \(n = 5\) nodes and \(m = 7\) edges.
  • Efficiency Calculation:
    • We substitute \(n=5\) and \(m=7\) into the formula: \[F(G) = 1 - \left[\frac{2(7 - 5 + 1)}{(5-1)(5-2)}\right] = 1 - \frac{6}{12} = 0.50\]
  • Interpretation:
    • With 5 nodes, a connected tree needs only 4 edges.
    • This graph has 3 redundant edges (\(7 - 5 + 1 = 3\)), resulting in a low efficiency score of \(0.50\).

Detailed Example: Graph 2 (More Efficient)

  • Graph 2 Structure:
    • Contains \(n = 7\) nodes and \(m = 10\) edges.
  • Efficiency Calculation:
    • We substitute \(n=7\) and \(m=10\) into the formula: \[F(G) = 1 - \left[\frac{2(10 - 7 + 1)}{(7-1)(7-2)}\right] = 1 - \frac{8}{30} \approx 0.73\]
  • Conclusion:
    • Despite having more edges (\(10\) vs \(7\)), Graph 2 is larger. Its redundancy is distributed over more nodes, yielding a higher efficiency of \(0.73\).
    • Graph 2 is more efficient than Graph 1 (\(0.73 > 0.50\)).

Graph Efficiency: Complete Graphs

  • Maximally Inefficient:
    • The complete graph (\(K_n\)) is minimally efficient because it has the maximum possible number of edges.
    • Number of edges: \(m = \frac{n(n-1)}{2}\)
  • Example (\(n=5\) Complete Graph):
    • Edges: \(m = \frac{5 \times 4}{2} = 10\)
    • Calculation: \[F(G) = 1 - \left[\frac{2(10 - 5 + 1)}{(5-1)(5-2)}\right] = 1 - \frac{12}{12} = 0.0\]
    • Thus, any complete graph yields \(F(G) = 0.0\).

Wrap-Up: Trees, Forests & Efficiency

  • Trees are minimally connected: exactly one path between any two nodes, every edge a bridge, and \(m = n - 1\).
  • Forests are what remains when a tree loses an edge — disjoint collections of trees.
  • Efficiency (Krackhardt, 1994) scores how economical connectivity is, from trees (\(F = 1\)) to complete graphs (\(F = 0\)).
  • Next up: Direction and reciprocity — metrics for directed graphs.