Longitudinal and Dynamic Social Networks - Revision history

Seth at 19:52, 28 October 2009

2009-10-28T19:52:10Z

Seth at 19:37, 28 October 2009

2009-10-28T19:37:06Z

Seth at 19:26, 28 October 2009

2009-10-28T19:26:27Z

New page

Wasserman/Faust says on p. 731
""Good, easy-to-use methods for longitudinal network data would be an important addition to the literature."

he also recommends:
Wasserman 1978
Iacobucci 1988
Iacobucci 1989,1990
Holland, Leinhardt 1981

More recently <bibtex>@article{snijders2005models,
title={{Models for longitudinal network data}},
author={Snijders, T.A.B.},
journal={Models and methods in social network analysis},
pages={215--247},
year={2005},
publisher={Cambridge Univ Pr}
}</bibtex>

says
"
This chapter treats statistical methods for network evolution. It
is argued that it is most fruitful to consider models where network
evolution is represented as the result of many (usually non-observed)
small changes occurring between the consecutively observed networks.
Accordingly, the focus is on models where a continuous-time network
evolution is assumed although the observations are made at discrete
time points (two or more).
Three models are considered in detail, all based on the assump-
tion that the observed networks are outcomes of a Markov process
evolving in continuous time. The independent arcs model is a trivial
baseline model. The reciprocity model expresses eﬀects of reciprocity,
but lacks other structural eﬀects. The actor-oriented model is based
on a model of actors changing their outgoing ties as a consequence of
myopic stochastic optimization of an ob jective function. This frame-
work oﬀers the ﬂexibility to represent a variety of network eﬀects. An
estimation algorithm is treated, based on a Markov chain Monte Carlo
implementation of the method of moments.
"

recommending:
"
Various models have been proposed for the statistical analysis of longitu-
dinal social network data. Earlier reviews were given by Wasserman (1978),
Frank (1991), and Snijders (1995).
"

@@ Line 58: / Line 58: @@
+"
-On the other side, actors’ characteristics – indicators of performance and
+On the other side, actors’ characteristics – indicators of performance and  success, attitudes and other cognitions, behavioral tendencies – can depend  on the social network the actor is situated in. It is well-known that in many  social situations, behavior and attitudes of individuals follow patterns of  assimilation to others to whom they are tied. Examples are the diﬀusion of  innovations in a professional community (Valente, 1995), pupils’ copying of  ‘chic’ behavior of their friends at school, or traders on a market copying the  allegedly successful behavior of their competitors.  The change of network structure is often referred to as selection (Lazars-  feld and Merton, 1954), and the change of individual characteristics of social  actors depending on the characteristics of others to whom they are tied is  called inﬂuence (Friedkin, 1998). It is assumed here that the group of actors  under study has been delineated in such a way that it is meaningful to inves-  tigate the selection and inﬂuence processes in this group without considering  ties to others outside the group. The necessity of studying selection and in-  ﬂuence processes in networks simultaneously was discussed both in detailed  network investigations (e.g., Padgett and Ansell, 1993) and in theoretical dis-  cussion essays (Emirbayer and Goodwin, 1994, Doreian and Stokman, 1997).  A concrete example is smoking initiation among adolescents, where it has  been established in the literature that friends tend to have similar patterns  of smoking behavior, but where it is unknown to which extent this is a mat-  ter of selection of friends on the basis of common behavior, or adaptation of  2 behavior towards that of one’s friends (Bauman and Ennett, 1996).  This chapter proposes a statistical method for investigating network struc-  ture together with relevant actor attributes as joint dependent variables in a  longitudinal framework, assuming that data have been collected according to  a panel design. This is a more detailed exposition of the proposals sketched  in Steglich et al. (2004). In the stochastic model, the network structure and  the individual attributes evolve simultaneously in a dynamic process. The  method is illustrated by an example on the dynamics of alcohol consumption  among adolescent friends.  ... The principles of actor-driven, or actor-oriented, modeling were proposed in Snijders (1996). The model for dynamics of only networks, without behav-  ior, was formulated in Snijders (2001, 2005). In Steglich et al. (2004), the sociological aspects of the model for dynamics of networks and behavior are discussed, with an extensive example about the interrelationship of the de- velopment of friendship networks and the dynamics in smoking and drinking behavior, on the basis of data from a Scottish high school. This chapter gives an overview of the speciﬁcation of the stochastic model for dynamics of networks and behavior and then proceeds to parameter estimation and model selection.
 success, attitudes and other cognitions, behavioral tendencies – can depend
 on the social network the actor is situated in. It is well-known that in many
 social situations, behavior and attitudes of individuals follow patterns of
-assimilation to others to whom they are tied. Examples are the diﬀusion of
+<bibtex>
-innovations in a professional community (Valente, 1995), pupils’ copying of
+@article{van2007actor,
-‘chic’ behavior of their friends at school, or traders on a market copying the
+  title={{An actor-oriented dynamic network approach: the case of interorganizational network evolution}},
-allegedly successful behavior of their competitors.
+  author={Van de Bunt, G.G. and Groenewegen, P.},
-The change of network structure is often referred to as selection (Lazars-
+  journal={Organizational Research Methods},
-feld and Merton, 1954), and the change of individual characteristics of social
+  volume={10},
-actors depending on the characteristics of others to whom they are tied is
+  number={3},
-called inﬂuence (Friedkin, 1998). It is assumed here that the group of actors
+  pages={463},
-under study has been delineated in such a way that it is meaningful to inves-
+  year={2007},
-tigate the selection and inﬂuence processes in this group without considering
+  publisher={Res Methods Div}
 ties to others outside the group. The necessity of studying selection and in-
 ﬂuence processes in networks simultaneously was discussed both in detailed
-network investigations (e.g., Padgett and Ansell, 1993) and in theoretical dis-
+</bibtex>
-cussion essays (Emirbayer and Goodwin, 1994, Doreian and Stokman, 1997).
+This is another really good lit review (they get better and better):
 A concrete example is smoking initiation among adolescents, where it has
-been established in the literature that friends tend to have similar patterns
+A  s demonstrated by three special issues in Journal of Mathematical Sociology (Doreian  & Stokman, 1996, 2003; Stokman & Doreian, 2001), the study of network dynamics  is of growing importance among sociologists, social psychologists, and network statisti-  cians. All three issues were devoted to the underlying mechanisms that induce the evolu-  tion of social networks: which micro mechanisms (i.e., individual choices) lead to which  macro outcomes (i.e., network structures), and how and why do these structures change  over time? More recently, research on networks and their dynamics is also ﬂourishing in  the strategy and organization literature; see, for instance, review articles by Borgatti and  Foster (2003) and Brass, Galaskiewicz, Greve, and Tsai (2004). One emerging and particu-  larly important topic is the ongoing dynamics of networks that result from collaborative  choices (among others, Ahuja, 2000; Chung, Singh, & Lee, 2000; Ebers, 1999; Gulati,  1995, 1999; Gulati & Gargiulo, 1999; Hagedoorn, 2006; Powell, 1998; Powell, Koput,  Smith-Doer, & Owen-Smith, 1999); today’s choice of an alliance partner affects tomorrow’s  options as it changes the network structure and thereby the future alternatives and strategies  of all fellow network members. For advancements in the latter topic, new statistical techni-  ques are required that can analyze these complex longitudinal network data structures (e.g.,  Hagedoorn, 2006). Most techniques, however, are not able to deal with these data struc-  tures in a satisfactorily and statistically sound way. Snijders (1995, 1996, 2001, 2005) and  Snijders & Van Duijn (1997) introduced actor-oriented modeling that does not assume  statistical independence between observations and combines continuous time Markov  analysis and random utility models. Originally these models were designed to model the  & Stokman, 1996, 2003; Stokman & Doreian, 2001), the study of network dynamics  is of growing importance among sociologists, social psychologists, and network statisti-  cians. All three issues were devoted to the underlying mechanisms that induce the evolu-  tion of social networks: which micro mechanisms (i.e., individual choices) lead to which  macro outcomes (i.e., network structures), and how and why do these structures change  over time? More recently, research on networks and their dynamics is also ﬂourishing in  the strategy and organization literature; see, for instance, review articles by Borgatti and  Foster (2003) and Brass, Galaskiewicz, Greve, and Tsai (2004). One emerging and particu-  larly important topic is the ongoing dynamics of networks that result from collaborative  choices (among others, Ahuja, 2000; Chung, Singh, & Lee, 2000; Ebers, 1999; Gulati,  1995, 1999; Gulati & Gargiulo, 1999; Hagedoorn, 2006; Powell, 1998; Powell, Koput,  Smith-Doer, & Owen-Smith, 1999); today’s choice of an alliance partner affects tomorrow’s  options as it changes the network structure and thereby the future alternatives and strategies  of all fellow network members. For advancements in the latter topic, new statistical techni-  ques are required that can analyze these complex longitudinal network data structures (e.g.,  Hagedoorn, 2006). Most techniques, however, are not able to deal with these data struc-  tures in a satisfactorily and statistically sound way. Snijders (1995, 1996, 2001, 2005) and  Snijders & Van Duijn (1997) introduced actor-oriented modeling that does not assume  statistical independence between observations and combines continuous time Markov  analysis and random utility models. Originally these models were designed to model the  evolution of expressive networks consisting of individuals (see van de Bunt, 1999; van de  Bunt, Van Duijn, & Snijders, 1999; van de Bunt, Wittek, & De Klepper, 2005; Van Duijn,  Zeggelink, Huisman, Stokman, & Wasseur, 2003). However, actor-oriented modeling is  also particularly adequate to capture the manner in which interﬁrm networks evolve as a  consequence of interaction between network structure and alliance partner choice. Actor-  oriented modeling is also an answer to a call for the application of agent-based modeling to  solve the puzzle of the interaction between network (i.e., macro) outcomes and a ﬁrm’s  (i.e., micro) choices (Macy & Willer, 2002). This is warranted because actor-oriented  modeling allows for assessing of the relevant mechanisms based on individual (i.e., ﬁrm)  choices. These choices are modeled under the assumption that ﬁrms are driven by the  expected amount of utility derived from the selection of speciﬁc partners, taking the present  evolution of expressive networks consisting of individuals (see van de Bunt, 1999; van de  Bunt, Van Duijn, & Snijders, 1999; van de Bunt, Wittek, & De Klepper, 2005; Van Duijn,  Zeggelink, Huisman, Stokman, & Wasseur, 2003). However, actor-oriented modeling is  also particularly adequate to capture the manner in which interﬁrm networks evolve as a  consequence of interaction between network structure and alliance partner choice. Actor-  oriented modeling is also an answer to a call for the application of agent-based modeling to  solve the puzzle of the interaction between network (i.e., macro) outcomes and a ﬁrm’s  (i.e., micro) choices (Macy & Willer, 2002). This is warranted because actor-oriented  modeling allows for assessing of the relevant mechanisms based on individual (i.e., ﬁrm)  choices. These choices are modeled under the assumption that ﬁrms are driven by the  expected amount of utility derived from the selection of speciﬁc partners, taking the present  network conﬁguration and the theoretically desired outcome explicitly into account.
 of smoking behavior, but where it is unknown to which extent this is a mat-
 ter of selection of friends on the basis of common behavior, or adaptation of
-−
 behavior towards that of one’s friends (Bauman and Ennett, 1996).
 This chapter proposes a statistical method for investigating network struc-
 ture together with relevant actor attributes as joint dependent variables in a
 longitudinal framework, assuming that data have been collected according to
 a panel design. This is a more detailed exposition of the proposals sketched
 in Steglich et al. (2004). In the stochastic model, the network structure and
 the individual attributes evolve simultaneously in a dynamic process. The
 method is illustrated by an example on the dynamics of alcohol consumption
 among adolescent friends.
 ...
 The principles of actor-driven, or actor-oriented, modeling were proposed in Snijders (1996). The model for dynamics of only networks, without behav-
 ior, was formulated in Snijders (2001, 2005). In Steglich et al. (2004), the sociological aspects of the model for dynamics of networks and behavior are discussed, with an extensive example about the interrelationship of the de- velopment of friendship networks and the dynamics in smoking and drinking behavior, on the basis of data from a Scottish high school. This chapter gives an overview of the speciﬁcation of the stochastic model for dynamics of networks and behavior and then proceeds to parameter estimation and model selection.
+"

@@ Line 19: / Line 19: @@
 says
+"
-This chapter treats statistical methods for network evolution. It
+ This chapter treats statistical methods for network evolution. It
 is argued that it is most fruitful to consider models where network
 evolution is represented as the result of many (usually non-observed)
@@ Line 26: / Line 26: @@
 evolution is assumed although the observations are made at discrete
 time points (two or more).
-Three models are considered in detail, all based on the assump-
+ Three models are considered in detail, all based on the assump-
 tion that the observed networks are outcomes of a Markov process
 evolving in continuous time. The independent arcs model is a trivial
@@ Line 40: / Line 40: @@
 recommending:
+"
-Various models have been proposed for the statistical analysis of longitu-
+ Various models have been proposed for the statistical analysis of longitu-
 dinal social network data. Earlier reviews were given by Wasserman (1978),
 Frank (1991), and Snijders (1995).
+"