| Terms |
Definitions |
|
80-20 rule
|
minority of the population is responsible for the majority of the activity- eg., sexual partners, friends, Heidi Roizen (silicon valley entrepreneur and CEO/board member) and Lois Weisberg(chicago connector- parties)
|
|
Closure
|
The probability of a tie between two nodes being formed increases as the number of common friends (groups) increases.Focal closure is where 2 people in an organization become friendsMembership closure is where a friend of some one become involved in an organization that the other is involved inTriadic Closure
|
|
Markov Chain Monte Carlo MLE
|
1.Start with guess for theta2.Create Network using ERGM simulations3.calculate log-ratio of liklihoog function4. Use Newton-Rhapson to find better theta5.Update theta, repeat 2-4
|
|
Row LAS
|
aggregated over the outlinks each actor claims to have (to whom they go for advice)
|
|
Cloud Computing
|
internet based computing whereby shared resources, software and information are provided to computers and other devices on-demand -Salesforce.com= Matt O'Connorpay as you go, multi-tennant, Auto upgrades, faster, cheaper, real-time collaboration, no software
|
|
Semantic Network
|
nodes are concepts and edges are relationships btw them-eg word association studies- Vannevar Bush show that humans have associative memory
|
|
Bow Ties
|
The model for the internet. The strongly connected component in the middle has sites that have out-links (link to it) but not in-links (are not linked to) and also a "out" part that is the opposite.There are tubes that can reach in and out pages but are not connected to the SCC. Also there are tendrils that can reach either the in or the out pages-also many employees depend on a small subset but not eachother
|
|
Theory of Transactive Memory
|
Good things happen to teams where1.members know (who knows)who knows what- reduces workload and redundancy2.high knowledge differentiation- expertise in different areas
|
|
Diffusion
|
Existing ties drive creation and destruction of attributesDepends on relative advantages, observability, compatibility with social system, Trialability (decrease risk by adopting gradually)- homophily can be barrier
|
|
Maximum Liklihood estimation
|
Choose theta such that the observed network configuration is the most likelygiven L=e^(g(y)*theta^T)/k(theta)
|
|
Theory of Proximity
|
A retrieves X from B if A is physically close to B, despite technology
|
|
Edge, stars, and triangle
|
star is an actor highly central to network
|
|
Info Cascades
|
herding/when beneficial to follow crowd despite own private info
|
|
T2K
|
text to knowledge is automatic data mining system
|
|
Clusters vs. Cascades
|
The spread of a new behavior can stall when it reaches a cluster (tightly-knit group)-homophily as a barrier to diffusion-cluster of density p is such that each node has p fraction of neighbors in setSet of all nodes is cluster density=1Union of 2 clusters with Density p has density pRunning into dense cluster is only thing that causes a cascade to stopComplete cascade only if no cluster with a critical density, q
|
|
Communication Network
|
people within a network (employees) who talk about (work) issues regularly-what is regularly?
|
|
homophily and assimilation
|
Homophily prevails among tobacco and alcohol users while assimilation prevails among canabis and alcohol users
|
|
Conformity Contagion
|
As more people do something there is more implicit social pressure to conform.-eg Milgram's Sky starers-in contrast to Rational Contagion (crowded bar vs empty bar) where something is contageous due to its info effects or direct-benefit effects
|
|
Correlation Coeff
|
The extent to which two variables are related/codetermined-indicates a predictive relationship
|
|
expansiveness bias
|
When survey respondents have different thresholds for a continuous variable with a non-continuous choice
|
|
MCMCMLE
|
Markov Chain Monte Carlo Maximum Liklihood Estimation-Simulation of graph distribution for given parameter values, refine values by comparison with observed graph-simple Markov inadequate when transitivity effects are strong
|
|
Assimilation
|
when nodes with friends take on the same attributes/join the same organizations
|
|
Selection vs Influence
|
Unchangeable characterists determine how links formed -vs. existing ties change alterable characteristice
|
|
Direct Benefit effects
|
Where copying others' actions has direct payoffs - network effects= where the value of something is proportional to the number of people that use it (Fax, Facebook)
|
|
ego bias
|
informants place themselves as more central in overall networks
|
|
Newton-Raphson Method
|
xsub(n+1)=xsub(n)-f'(xsub(n))/f"(xsub(n))
|
|
Theory of social exchange
|
If A gets info X from B, then B is more likely to get info Y from A
|
|
3 Rules of Epidemics
|
1. Stickiness- the message must make an impact- the disease must be powerful- syphilis occurred in more latent form earlier in century but didnt become epidemic2. Law of Few- power laws- few very sexually active people get disease cause it to become epidemic3. Power of Context- how alter behavior based on context- people are very sensitive to the environment- stabbing of Kitty Genovese with 30 wittnesses-mid 1990s Baltimore houses razed and cutback on medical spending lead to syphilisepidemic
|
|
Name interpreter
|
obtain info on the alters and their relationships
|
|
Bystander Effect
|
Where individuals do not offer help in an emergency when other are present. From the power of context and Kitty Genovese where thirty wittnesses observed her getting mugged and stabbed but expected the other people to call the cops then no one did
|
|
Theories of Contagion
|
A retrieves info on X from B if others in A's comm network also do so
|
|
ego/alter effect
|
feedback effects in the complex process of Influence including diffusion and coevolution
|
|
no Homophily vs heterophily
|
Random ties vs ties more likely between boys and girls (different attributes)
|
|
Component
|
Connected subset of network nodes and links
|
|
single v multiple name generators
|
Single elicits small network of close ties- can use social exchange criteria (advice), some use affective criteria (closeness, role (neighbor), frequency of interaction-multiple ask many questions (eg., who socialize with, advice...)
|
|
Data Collection Methods
|
Surveys- collect perception of interactionlist of names v free recallRatings v complete rankings-Observations- face-face, listserv...-Interviews- face-face, telephone, Snowball Sampling-Indirect data- archival records- more reliable-Experiments
|
|
Disciplinary Science Model
|
where Journals in one disclipline cite journals in another- shows how disciplines are linked
|
|
Column LAS
|
Captures which inlinks each actor reports to have (who trusts them).-This shows whether people feel needed/involved in a network (bc ppl come to them to talk/get advice)
|
|
ERGM/p* models
|
give info about distribution of complex social interactions2.see if network structures are observed more than expected by chance3.Allow for quantitative modeling 4.break micro-macro gap - each tie is random variable-Propose dependence hypothesis, implies particular model form-Markov random graphs are a class of ERGM with Conditional Dependence assumption
|
|
MTML
|
Multitheoretical Multilevel analysis refers to the approach that explains the creation, maintenance, and dissolution of network links through different theories and at different analytical levels
|
|
Branching Process
|
Simplest model of ContagionFirst wave(=k)- carrier meets k people, some infectedSecond wave(=k*k people)- 1st wave meets k new people and pass disease with prob=pHas basic reproductive number- expected number of people infected by an individual (R=pk)>1 necessary for disease to not die out
|
|
(Theory of) Co-evolution
|
Why attributes change based on links and why links change based on attributes-ex Substance abuse, sports, and friendshipWhen friends have similar (alterable) characteristics because they develop them together
|
|
Strongly Connected Component
|
A subset of nodes for each of which there exists a directed path to every other node in the SCC
|
|
Theory of Collective Action vs mutual interest
|
Collective action is where actors work together to create something that they all enjoy-MI- Actors work together to each get something (eg., GroupOn)
|
|
QAP
|
Quadratic Assignment Procedure- sometimes used as goodness of fit test for graph-level statistics using Monte Carlo MLE
|
|
Theory of Homophily
|
A is more likely to retrieve X from B if A and B share attributes (position, gender, etc.)- if % cross gender edges>> 2*%male in network*%female in network
|
|
Name Generator
|
- identify an ego's alters-typically ego-centric studies that set boundaries during study
|
|
Cognitive Social Structure
|
Everyone in a network's perceptions of what links exist in a network-1987 Krackhardt- CSS have more info than normal social structure, no "objective relations
|
|
Expanding selection/snowball sampling
|
Start with initial list of people then add based on their responses- could then select k-core (know all except k members in set)
|
|
Component
|
Connected subset of network nodes and links
|
|
Roger's adoption Process
|
KnowledgePersuasiondecisionimplementationConfirmation
|
|
Setting Network Boundaries
|
1. Positional (eg employment)2. Event Based (who attended)3. Relational (social interconnectedness)
|
|
Informant inaccuracy
|
BKS performed a study in which they observed who people on the beach were talking to and then asked them to report who they were talking to- different
|
|
Reverse small world method
|
to determine network size- out of 500 people given their jobs, an individual has to say one person they know who would have the highest probability of knowing that person
|
|
Theory of balance
|
ties are more likely to form to create a balanced network
|
