롤 데이터 분석 연구의 일환으로 공부한 내용 정리
1. Size
- 150 is an optimal size for a human group
- 280 is the average size of acquaintance networks in U.S.
2. Density
- the number of connections in the network
- it is a key attribute of a network and should always be included as a covariate in analyses testing the effects of other structural properties
-> because there is an inverse relationship between size and density.
[asymmetric networks] - directed
- D = L / N(N-1)
-> L=the number of links, N=network size
[symmetric networks] - undirected
- D = L*2 / N(N-1)
3. Mutuality / Reciprocity
- reciprocated tied are links in which the direction goes both ways
-> 롤 연구에서 측정 불가능
4. Triads / Transitivity
- the links between three nodes
-> x
5. Diameter / Average Path Length (APL)
- diameter is the length of the longest path in the network (=number of steps)
- low diameter or low APL indicate a cohesive network with little clustering
- high diameter and low APL indicates that there are branches or spurs to the network that are mostly inaccessible to others in the network.
- generate a sample network with the same size and density and compare with the original network.
6. Density and Cohesion
- As density increases, APL decreases
- density values above 50% indicate networks that have many redundant paths between nodes.
-> do not contain much structual information and in many cases are not likely to be interesting subtantively.
-> in such cases, it will probably be desirable to prune the network so that the researcher can find the hidden structure within the tangle of links.
-> pruning 기법 조사하기
7. Clustering
- it is calculated by comupting the average of each node's personal network density.
- another way to calculate clustering is to calculate the ratio of closed triads to all podsible triads in the network.
- a network with high average personal network densities indicates that people connected to a third person tend to be connected to each other.
- when personal network densities are low, transivity is lacking and people are not more likely to know each other when connected to the same third party.
- 2 different types of network structures exist that link clustered subgroups.
1) a bridge structure in which the clustered subgroups are linked by bridges
2) a centralized structure in which the sub-groups are linked by central nodes who occupy powerful positions in the network.
8. Centralization
- it is the degree a network's ties are focused on one person or a set of people.
- dendity and centralization are not correlated and are independent structural measures of the network.
- centrality = the number of links a person sends and receives
- centralization degress (CD) = (SUM(Max(centrality) - centrality)) / n^2 -3n + 2
-> n = network size
9. Core-Periphery
Network-Level Measures