The Statistical Analysis of the Dynamics of Networks and Behaviour. An Introduction to the Actor-based Approach. Christian Steglich and Tom Snijders ——————

Slides:



Advertisements
Similar presentations
Introduction Describe what panel data is and the reasons for using it in this format Assess the importance of fixed and random effects Examine the Hausman.
Advertisements

Statistical Social Network Analysis - Stochastic Actor Oriented Models Johan Koskinen The Social Statistics Discipline Area, School of Social Sciences.
Analysing network-behavioural co-evolution with SIENA Christian SteglichUniversity of Groningen Tom SnijdersUniversity of Groningen Mike PearsonNapier.
Mobile Communication Networks Vahid Mirjalili Department of Mechanical Engineering Department of Biochemistry & Molecular Biology.
Generalized Method of Moments: Introduction
CHAPTER 15: Tests of Significance: The Basics Lecture PowerPoint Slides The Basic Practice of Statistics 6 th Edition Moore / Notz / Fligner.
Assessing influence and selection in network-behavioural co-evolution with an application to smoking and alcohol consumption among adolescents. Christian.
Probabilistic Reasoning over Time
Network Matrix and Graph. Network Size Network size – a number of actors (nodes) in a network, usually denoted as k or n Size is critical for the structure.
Bayesian Network and Influence Diagram A Guide to Construction And Analysis.
Where we are Node level metrics Group level metrics Visualization
Some results from Scottish data Topic smoking behaviour and friendship Problem influence and/or selection Theory drifting smoke rings (Pearson, West, Michell)
Advanced model specification for SIENA models: score tests and forward model selection Mark HuismanUniversity of Groningen Christian SteglichUniversity.
Introduction to VISSIM
Categorisation of decision situations affecting induced preferences. Some empirical tests of a formal framing model. Dr. Christian Steglich ICS / department.
An Introduction to Variational Methods for Graphical Models.
Introduction of Probabilistic Reasoning and Bayesian Networks
Ai in game programming it university of copenhagen Statistical Learning Methods Marco Loog.
Link creation and profile alignment in the aNobii social network Luca Maria Aiello et al. Social Computing Feb 2014 Hyewon Lim.
Chapter 4 Validity.
Statistical NLP: Lecture 11
Analysing the co-evolution of social networks and “behavioural” dimensions with SIENA Christian SteglichUniversity of Groningen Tom SnijdersUniversity.
Exponential random graph (p*) models for social networks Workshop Harvard University February 2002 Philippa Pattison Garry Robins Department of Psychology.
Some results from Scottish data The Statistical Analysis of the Dynamics of Networks and Behaviour: An Application to Smoking and Drinking Behaviour among.
Co-evolution of Members’ Attachment to the Team and Team Interpersonal Networks Chunke Su Noshir Contractor University of Illinois at Urbana-Champaign.
Networks, Lie Monoids, & Generalized Entropy Metrics Networks, Lie Monoids, & Generalized Entropy Metrics St. Petersburg Russia September 25, 2005 Joseph.
AGC DSP AGC DSP Professor A G Constantinides© Estimation Theory We seek to determine from a set of data, a set of parameters such that their values would.
Analysing the co-evolution of social networks and “behavioural” dimensions with SIENA Christian SteglichUniversity of Groningen Tom SnijdersUniversity.
2. Point and interval estimation Introduction Properties of estimators Finite sample size Asymptotic properties Construction methods Method of moments.
Using ranking and DCE data to value health states on the QALY scale using conventional and Bayesian methods Theresa Cain.
Sunbelt 2009statnet Development Team ERGM introduction 1 Exponential Random Graph Models Statnet Development Team Mark Handcock (UW) Martina.
7-1 Introduction The field of statistical inference consists of those methods used to make decisions or to draw conclusions about a population. These.
 We cannot use a two-sample t-test for paired data because paired data come from samples that are not independently chosen. If we know the data are paired,
Copyright © Cengage Learning. All rights reserved. 11 Applications of Chi-Square.
Using Friendship Ties and Family Circles for Link Prediction Elena Zheleva, Lise Getoor, Jennifer Golbeck, Ugur Kuter (SNAKDD 2008)
6 am 11 am 5 pm Fig. 5: Population density estimates using the aggregated Markov chains. Colour scale represents people per km. Population Activity Estimation.
Object-Oriented Software Engineering Practical Software Development using UML and Java Chapter 8: Modelling Interactions and Behaviour.
Exploring the dynamics of social networks Aleksandar Tomašević University of Novi Sad, Faculty of Philosophy, Department of Sociology
Modeling & Simulation: An Introduction Some slides in this presentation have been copyrighted to Dr. Amr Elmougy.
Principles of Social Network Analysis. Definition of Social Networks “A social network is a set of actors that may have relationships with one another”
Bayesian networks Classification, segmentation, time series prediction and more. Website: Twitter:
Various topics Petter Mostad Overview Epidemiology Study types / data types Econometrics Time series data More about sampling –Estimation.
CHAPTER 17: Tests of Significance: The Basics
1.3 Simulations and Experimental Probability (Textbook Section 4.1)
Economics 173 Business Statistics Lecture 4 Fall, 2001 Professor J. Petry
Processing Sequential Sensor Data The “John Krumm perspective” Thomas Plötz November 29 th, 2011.
A Passive Approach to Sensor Network Localization Rahul Biswas and Sebastian Thrun International Conference on Intelligent Robots and Systems 2004 Presented.
CHAPTER 15: Tests of Significance The Basics ESSENTIAL STATISTICS Second Edition David S. Moore, William I. Notz, and Michael A. Fligner Lecture Presentation.
Computer Vision Lecture 6. Probabilistic Methods in Segmentation.
Lecture PowerPoint Slides Basic Practice of Statistics 7 th Edition.
POSC 202A: Lecture 4 Probability. We begin with the basics of probability and then move on to expected value. Understanding probability is important because.
AP Statistics Section 11.1 B More on Significance Tests.
Review of fundamental 1 Data mining in 1D: curve fitting by LLS Approximation-generalization tradeoff First homework assignment.
Introduction to Statistical Models for longitudinal network data Stochastic actor-based models Kayo Fujimoto, Ph.D.
CHAPTER 15: Tests of Significance The Basics ESSENTIAL STATISTICS Second Edition David S. Moore, William I. Notz, and Michael A. Fligner Lecture Presentation.
Introduction to ERGM/p* model Kayo Fujimoto, Ph.D. Based on presentation slides by Nosh Contractor and Mengxiao Zhu.
Does the brain compute confidence estimates about decisions?
+ Testing a Claim Significance Tests: The Basics.
Unit 5: Hypothesis Testing
Dr.MUSTAQUE AHMED MBBS,MD(COMMUNITY MEDICINE), FELLOWSHIP IN HIV/AIDS
Adoption of Health Information Exchanges and Physicians’ Referral Patterns: Are they Mutually Reinforcing? SAEEDE EFTEKHARI*, School of Management, State.
Overview and Basics of Hypothesis Testing
CHAPTER 21: Comparing Two Means
Chapter 4. Probability
Agent-Based Methods for Dynamic Social Networks
Amblard F.*, Deffuant G.*, Weisbuch G.** *Cemagref-LISC **ENS-LPS
Significance Tests: The Basics
Basic Practice of Statistics - 3rd Edition Introduction to Inference
Probability and Time: Markov Models
Longitudinal Social Network Data
Presentation transcript:

The Statistical Analysis of the Dynamics of Networks and Behaviour. An Introduction to the Actor-based Approach. Christian Steglich and Tom Snijders —————— 2003/04.

Situation investigated: Given is a group of actors i  {1,…,N}, - this group is ‘carrier’ of a meaningful social network x, and - actors in this group show behaviour z. Behaviour and network positions of actors are interdependent. Problem investigated: How does this interdependence come into existence?  What are the dynamic mechanisms generating network ties and behaviour?

Black actor reciprocates friendship White actor adapts to (perceived) friend Selection mechanisms lead to changes in network ties: Influence mechanisms lead to changes in actor characteristics: Two broad types of mechanisms that drive such co-evolution:

Black actor reciprocates friendship Selection mechanism followed by influence mechanism: Influence mechanism followed by selection mechanism: White actor adapts to (per- ceived) friend White actor adapts to (re- ciprocal) friend Black actor reciprocates friendship Both types of mechanisms can occur in the same process:

Black actor reciprocates friendship White actor adapts to (per- ceived) friend White actor adapts to (re- ciprocal) friend Black actor reciprocates friendship Problem: Due to sparse data, in many cases the order of occurrence of these mechanisms cannot be identified… When working with panel data, dynamics between measurements are not known.

Black actor reciprocates friendship White actor adapts to (per- ceived) friend White actor adapts to (re- ciprocal) friend Black actor reciprocates friendship Problem: …but in many cases this order of occurrence is of focal interest from the theory perspective. Theory A: Relationships are governed by norms of reciprocity. Adaptive behaviour occurs most likely within close (reciprocated) relationships.

Black actor reciprocates friendship White actor adapts to (per- ceived) friend White actor adapts to (re- ciprocal) friend Black actor reciprocates friendship Problem: …but in many cases this order of occurrence is of focal interest from the theory perspective. Theory B: Influence is strongest in asymmetrical relationships. Homophily is a strong deter- minant of starting a new relationship.

Theory B: Influence is strongest in asymmetrical relationships. Homophily is a strong deter- minant of starting a new relationship. Theory A: Relationships are governed by norms of reciprocity. Adaptive behaviour occurs most likely within close (reciprocated) relationships. ?

How to test such theories against each other? longitudinal data (we will be studying panel data), explicit modelling of the mechanisms driving co-evolution, fit model to data, infer relative strength of the different mechanisms from parameter estimates, draw conclusions about the theories, based on evidence for the mechanisms they postulate.

Continuous time Markov process model: state space consists of all possible configurations of network ties and behaviourals, individual decisions modelled by objective functions: – one for behavioural change (in ‘micro steps’), – another one for network change (in ‘micro steps’); timing of individual decisions by rate functions: – again one for behavioural decisions, – and another one for network decisions.

State space Pair (x,z)(t) contains … – adjacency matrix x and – vector(s) of behaviourals z – at time point t. Co-evolution is modelled by specifying transition probabilities between such states (x,z)(t 1 ) and (x,z)(t 2 ).

16 possible states for a network consisting of one dyad only. (assuming actor characteristics and network ties to be dichotomous)

n # K32M64G512T16E For the simplest case of dichotomous ties and one dichotomous actor characteristic, the cardinality of the state space increases quickly with the number of actors: Some numbers for illustration:

Transitions between states: Not all possible transitions (x,z)(t 1 )  (x,z)(t 2 ) are modelled, but only “micro steps” are: – network micro step: (x,z)(t 1 ) and (x,z)(t 2 ) differ in one tie x ij only. –behavioural micro step: (x,z)(t 1 ) and (x,z)(t 2 ) differ in one behavioural score z i only. Observed transitions are more complex -- they are inter- preted as resulting from a sequence of such micro steps.

Possible changes of network ties: (diagram renders possible network micro steps only)

Possible changes of behaviourals: (diagram renders possible behavioural micro steps only)

All possible micro- transitions for a one-dyad network:

Actor based modelling: The modelled transitions (x,z)(t 1 )  (x,z)(t 2 ) are results of individual decision making. – network micro step: actor i maximises “value of his network- behavioural neighbourhood” by changing tie to actor j. – behavioural micro step: actor i maximises a similar “value of his network- behavioural neighbourhood” by changing his behavioural score.

Actor based modelling: The “value of network-behavioural neighbourhood” is operationalised by satisfaction measures: – satisfaction of actor i from changing the network tie to actor j: f deterministic satisfaction measure,  random distortion of convenient choice. – similar (but separate) model for satisfaction with behavioural decisions.

Actor based modelling: The deterministic part f of the satisfaction measure consists of the following components: – a function measuring utility (based only on resulting network configuration), – a function measuring endowment effects (based on current and resulting network), – a function measuring reinforcement learning (also based on current and resulting network).

Actor based modelling: The probabilistic part  of the satisfaction measure is chosen as i.i.d. of extreme value type I : – this way, the choice probabilities can be expressed as (for network decisions, behavioural decisions analogous).

Changes under control of the upper-left actor: red transitions are behavioural changes, green transitions are network changes.

Changes under control of the lower-right actor: (same colouring) One can see that an individual actor’s scope of action is rela- tively small.

Interpretation of parameter estimates: Rate function parameters indicate the speed of the respective evolution process. – positive parameter attached to an effect means quicker changes in the process when the effect is present. Objective function parameters indicate the actor’s preferences. – positive parameter attached to an effect means a higher preference of the actor for a decision in which the effect is present. Nota bene: parameter estimates do NOT indicate the network-behavioural co-evolution from a macro perspective!

Remarks on model estimation: The likelihood of an observed data set cannot be calculated in closed form, but can at least be simulated.  ‘third generation problem’ of statistical analysis,  simulation-based inference is necessary. Currently available: – Method of Moments estimation (Snijders 2001, 1998) – Maximum likelihood approach (Snijders & Koskinen 2003) Implementation: program SIENA, part of the StOCNet software package.