《Pajek社会网络探索性分析》书籍简介

Exploratory Social Network Analysis with Pajek

Part I Fundamentals

Social network analysis focuses on ties among, for example, people, groups of people, organizations, and countries. These ties combine to form networks, which we will learn to analyze. The first part of this book introduces the concept of a social network. We discuss several types of networks and the ways in which we can analyze them numerically and visually with the computer software program Pajek, which is used throughout this book. After studying Chapters 1 and 2, you should understand the concept of a social network and be able to create, manipulate, and visualize a social network with the software presented in this book.

社交网络分析侧重于例如人、人群、组织和国家之间的联系。 这些联系结合起来形成网络，我们将学习分析。 本书的第一部分介绍了社交网络的概念。 我们讨论了几种类型的网络以及我们可以使用计算机软件程序 Pajek 对它们进行数值和可视化分析的方法，Pajek 贯穿本书。 学习第 1 章和第 2 章后，您应该了解社交网络的概念，并能够使用本书中提供的软件创建、操作和可视化社交网络。

Chapter 1 Looking for Social Structure

The social sciences focus on structure: the structure of human groups, communities, organizations, markets, society, or the world system. In this book, we conceptualize social structure as a network of social ties. Social network analysts assume that interpersonal ties matter, as do ties among organizations or countries, because they transmit behavior, attitudes, information, or goods. Social network analysis offers the methodology to analyze social relations; it tells us how to conceptualize social networks and how to analyze them.

社会科学关注结构：人类群体、社区、组织、市场、社会或世界体系的结构。 在本书中，我们将社会结构概念化为一个社会关系网络。 社交网络分析师认为人际关系很重要，就像组织或国家之间的联系一样，因为它们传递行为、态度、信息或商品。 社交网络分析提供了分析社会关系的方法； 它告诉我们如何概念化社交网络以及如何分析它们。

In this book, we present the most important methods of exploring social networks, emphasizing visual exploration. Network visualization has been an important tool for researchers from the very beginning of social network analysis. This chapter introduces the basic elements of a social network and shows how to construct and draw a social network.

在本书中，我们展示了探索社交网络的最重要方法，强调视觉探索。 从社交网络分析的一开始，网络可视化就一直是研究人员的重要工具。 本章介绍了社交网络的基本要素，并展示了如何构建和绘制社交网络。

Chapter 2 Attributes and Relations

In Chapter 1, we argued that social network analysis focuses on social relations. A network is a set of vertices and lines. Both vertices and lines have characteristics that we may want to include in our analysis (e.g., the gender of people and the strength of their ties). As noted in Chapter 1, properties of relations are represented by line values in the network (e.g., first and second choices among girls in the dormitory). Now, we add characteristics of the vertices to the analysis. How can we use information on the actors to make sense of the social network?

在第 1 章中，我们认为社会网络分析侧重于社会关系。 网络是一组顶点和线。 顶点和线都具有我们可能希望包含在我们的分析中的特征（例如，人的性别和他们的联系强度）。 如第 1 章所述，关系的属性由网络中的线值表示（例如，宿舍中女孩的第一和第二选择）。 现在，我们将顶点的特征添加到分析中。 我们如何使用有关演员的信息来理解社交网络？

In this chapter, we present techniques that combine relational network data and nonrelational attributes, such as psychological, social, economic, and geographical characteristics of the vertices in the network. The attributes enhance our interpretation of network structure, and they enable us to study subsections of the network. In addition, we briefly discuss how to use the network position of vertices in statistical analysis; social network analysis and statistics are two complementary sets of techniques. After having studied this chapter, you will understand the basic data used in network analysis and will be able to combine relational and nonrelational data.

在本章中，我们将介绍将关系网络数据和非关系属性（例如网络中顶点的心理、社会、经济和地理特征）相结合的技术。 这些属性增强了我们对网络结构的解释，并使我们能够研究网络的子部分。 此外，我们简要讨论了如何在统计分析中使用顶点的网络位置； 社交网络分析和统计是两组互补的技术。 学习本章后，您将了解网络分析中使用的基本数据，并能够将关系数据和非关系数据结合起来。

Part II Cohesion

Solidarity, shared norms, identity, collective behavior, and social cohesion are considered to emerge from social relations. Therefore, the first concern of social network analysis is to investigate who is related and who is not. Why are some people or organizations related, whereas others are not? The general hypothesis here states that people who match on social characteristics will interact more often, and people who interact regularly will foster a common attitude or identity.

团结、共享规范、身份、集体行为和社会凝聚力被认为是从社会关系中产生的。 因此，社交网络分析的首要关注点是调查谁相关，谁不相关。 为什么有些人或组织是相关的，而有些则不是？ 这里的一般假设表明，社会特征匹配的人会更频繁地互动，而经常互动的人会培养共同的态度或身份。

In this part of the book, which covers Chapters 3 through 5, we discuss several measures of cohesion. You will learn to detect cohesive subgroups within several types of social networks.

在本书的这一部分（涵盖第 3 章到第 5 章）中，我们讨论了几种凝聚力的衡量标准。 您将学习在几种类型的社交网络中检测有凝聚力的子群。

3 Cohesive Subgroups

Social networks usually contain dense pockets of people who “stick together.” We call them cohesive subgroups, and we hypothesize that the people involved are joined by more than interaction. Social interaction is the basis for solidarity, shared norms, identity, and collective behavior, so people who interact intensively are likely to consider themselves a social group. Perceived similarity, for instance, membership of a social group, is expected to promote interaction. We expect similar people to interact a lot, at least more often than with dissimilar people. This phenomenon is called homophily or assortativity: Birds of a feather flock together. We will learn how to measure this phenomenon in Chapter 6.

社交网络通常包含大量“团结在一起”的人。我们称他们为有凝聚力的子群体，我们假设所涉及的人不仅仅是通过互动来加入的。社会互动是团结、共享规范、身份和集体行为的基础，因此互动频繁的人很可能将自己视为一个社会群体。感知的相似性，例如社会团体的成员身份，有望促进互动。我们希望相似的人进行很多互动，至少比与不同的人互动更频繁。这种现象被称为同质性或分类性：羽毛般的鸟类聚集在一起。我们将在第 6 章学习如何测量这种现象。

In this chapter, we present a number of techniques to detect cohesive subgroups in social networks, all of which are based on the ways in which vertices are interconnected. These techniques are a means to an end rather than an end in themselves. The ultimate goal is to test whether structurally delineated subgroups differ with respect to other social characteristics, for instance, norms, behavior, or identity. Does the homophily principle work? May we conclude that a cohesive subgroup represents an emergent or an established social group?

在本章中，我们介绍了许多用于检测社交网络中的内聚子组的技术，所有这些技术都基于顶点互连的方式。这些技术本身是达到目的的手段，而不是目的。最终目标是测试结构上描绘的亚群是否在其他社会特征方面有所不同，例如规范、行为或身份。同质性原则有效吗？我们可以得出结论，一个有凝聚力的子群体代表一个新兴的或一个成熟的社会群体吗？

4 Sentiments and Friendship

In the preceding chapter, we discussed several techniques for finding cohesive subgroups within a social network. People who belong together tend to interact more frequently than people who do not. In the current chapter, we extend this idea to affective relations that are either positive or negative, for instance, friendship versus hostility, liking versus disliking. We expect positive ties to occur within subgroups and negative ties between subgroups.

在前一章中，我们讨论了在社交网络中寻找有凝聚力的子群的几种技术。 属于一起的人往往比不属于一起的人更频繁地互动。 在本章中，我们将这个想法扩展到积极或消极的情感关系，例如，友谊与敌意、喜欢与不喜欢。 我们预计子组内会出现正联系，而子组之间会出现负联系。

Hypotheses about patterns of affective relations stem from social psychology, and they are widely known as balance theory. First, we introduce this theory and discuss how it was incorporated in network analysis. Then, we apply it to affective relations, that is, social relations that are subjective and mental rather than tangible.

关于情感关系模式的假设源于社会心理学，它们被广泛称为平衡理论。 首先，我们介绍这个理论并讨论它是如何被纳入网络分析的。 然后，我们将其应用于情感关系，即主观和心理而非有形的社会关系。

5 Affiliations

Membership of an organization or participation in an event is a source of social ties. In organizations and events, people gather because they have similar tasks or interests and they are likely to interact. Members of a sports club, for instance, share a preference for a particular sport and play with or against one another. Directors and commissioners on the boards of a corporation are collectively responsible for its financial success and meet regularly to discuss business matters. Inspired by the sociology of Georg Simmel, groups of people who gather around one or more organizations and events are called social circles.

组织的成员资格或参加活动是社会联系的来源。在组织和活动中，人们聚集是因为他们有相似的任务或兴趣，并且很可能会互动。例如，体育俱乐部的成员对一项特定的运动有着共同的偏好，并相互比赛或对抗。公司董事会的董事和委员共同负责其财务成功，并定期开会讨论业务问题。受 Georg Simmel 社会学的启发，聚集在一个或多个组织和事件周围的人群被称为社交圈。

In previous chapters, we studied direct ties among people, such as the choice of friends, or among other social entities, for instance, trade relations between countries. Note that we studied relations among actors of one kind: relations between people or between organizations but not between people and organizations. Now, we focus on the latter type, which is called an affiliation. Data on affiliations can be obtained relatively easily, and they are very popular in data mining.

在前面的章节中，我们研究了人与人之间的直接联系，例如朋友的选择，或其他社会实体之间的直接联系，例如国家之间的贸易关系。请注意，我们研究了一种行为者之间的关系：人与人之间或组织之间的关系，而不是人与组织之间的关系。现在，我们关注后一种类型，称为从属关系。隶属关系的数据可以相对容易地获得，并且它们在数据挖掘中非常流行。

Affiliations are often institutional or “structural,” that is, forced by circumstances. They are less personal and result from private choices to a lesser degree than sentiments and friendship. Of course, membership in a sports team depends much more on a person’s preferences than detention in a particular prison ward, but even the composition of sports teams depends on circumstances and on decisions made by coaches and sports club authorities. Affiliations express institutional arrangements; and because institutions shape the structure of society, networks of affiliations tell us a lot about society. People are often affiliated to several organizations and events at the same time, so they belong with a number of social circles; or, in other words, they are the intersection of many social circles. Society may be seen as a fabric of intersecting social circles.

隶属关系通常是制度性的或“结构性的”，即受环境所迫。与情感和友谊相比，它们不那么个人化，并且是私人选择的结果。当然，运动队的成员资格更多地取决于个人的喜好，而不是被拘留在特定的监狱病房，但即使是运动队的组成也取决于情况以及教练和体育俱乐部当局的决定。隶属关系明示制度安排；因为制度塑造了社会的结构，从属关系网络告诉我们很多关于社会的信息。人们往往同时隶属于多个组织和事件，因此他们属于多个社交圈；或者说，他们是很多社交圈的交集。社会可以被视为交叉社交圈的结构。

Although membership lists do not tell us exactly which people interact, communicate, and like each other, we may assume that there is a fair chance that they will. Moreover, joint membership in an organization often entails similarities in other social domains. If, for example, people have chosen to become members in (or have been admitted to) a particular golf club, they may well have similar professions, interests, and social status. Different types of affiliations do not overlap in a random manner: Social circles usually contain people who are clustered by affiliations to more than one type of organization. From the number or intensity of shared events, we may infer the degree of similarity of people. However, this argument can be reversed: Organizations or events that share more members are also closer socially. A country club with many members from the local business elite can be said to be part of the business sphere.

尽管成员名单并不能准确告诉我们哪些人会互动、交流和喜欢彼此，但我们可以假设他们很有可能会这样做。此外，一个组织中的联合成员身份通常需要其他社会领域的相似性。例如，如果人们选择成为（或已被接纳）某个高尔夫球俱乐部的会员，他们很可能具有相似的职业、兴趣和社会地位。不同类型的隶属关系不会以随机方式重叠：社交圈通常包含通过隶属关系聚集到一种以上组织类型的人。从共享事件的数量或强度，我们可以推断出人们的相似程度。然而，这个论点可以颠倒：共享更多成员的组织或事件在社交上也更接近。一个拥有众多本地商界精英成员的乡村俱乐部，可以说是商界的一部分。

In this chapter, we present a technique for analyzing networks of affiliations that focuses on line values. In addition, we discuss three-dimensional displays of social networks.

在本章中，我们将介绍一种分析从属网络的技术，该技术侧重于线值。此外，我们还讨论了社交网络的三维显示。

Part III Brokerage

In quite a few theories, social relations are considered channels that transport information, services, or goods between people or organizations. In this perspective, social structure helps to explain how information, goods, or even attitudes and behavior diffuse within a social system. Network analysis reveals social structure and helps to trace the routes that goods and information may follow. Some social structures permit rapid diffusion of information, whereas others contain sections that are difficult to reach.

在很多理论中，社会关系被认为是在人或组织之间传输信息、服务或商品的渠道。从这个角度来看，社会结构有助于解释信息、商品，甚至态度和行为如何在社会系统中扩散。网络分析揭示了社会结构，并有助于追踪商品和信息可能遵循的路线。一些社会结构允许信息快速传播，而另一些则包含难以到达的部分。

This is a bird’s-eye view of an entire social network. However, we can also focus on the position of specific people or organizations within the network. In general, being well connected is advantageous. Contacts are necessary to have access to information and help. The number and intensity of a person’s ties are called his or her sociability or social capital, which is known to correlate positively with age and education in West- ern societies. Some people occupy central or strategic positions within the system of channels and are crucial for the transmission process. Such positions may put pressure on their occupants, but they may also yield power and profit.

这是整个社交网络的鸟瞰图。但是，我们也可以关注特定人或组织在网络中的位置。一般来说，连接良好是有利的。联系人是获取信息和帮助所必需的。一个人的联系的数量和强度被称为他或她的社交能力或社会资本，众所周知，在西方社会，这与年龄和教育呈正相关。有些人在渠道系统中占据中心或战略位置，对传播过程至关重要。这样的职位可能会给他们的居住者带来压力，但他们也可能会产生权力和利润。

In this part of the book, we focus on social networks as structures that allow for the exchange of information. In this approach, the direction of ties is not very important, so we discuss only undirected networks (with one exception). In Chapter 6, we present the concepts of centrality and centralization. In Chapter 7, we discuss the structure of the immediate network of actors, especially the pressure or power that is connected to particular structures of this ego network. In Chapter 8, we take time into account as we study the role of network structure in the diffusion of innovations and diseases.

在本书的这一部分，我们将社交网络作为允许信息交换的结构。在这种方法中，关系的方向不是很重要，所以我们只讨论无向网络（有一个例外）。在第 6 章中，我们介绍了中心性和中心化的概念。在第 7 章中，我们讨论了行动者的直接网络的结构，尤其是与这个自我网络的特定结构相关的压力或权力。在第 8 章中，我们在研究网络结构在创新和疾病传播中的作用时考虑了时间。

6 Center and Periphery

In this chapter, we present the concepts of centrality and centralization, which are two of the oldest concepts in network analysis. Most social networks contain people or organizations that are central. Because of their position, they have better access to information and better opportunities to spread information. This is known as the ego-centered approach to centrality. Viewed from a sociocentered perspective, the network as a whole is more or less centralized. Note that we use centrality to refer to positions of individual vertices within the network, whereas we use centralization to characterize an entire network. A network is highly centralized if there is a clear boundary between the center and the periphery. In a highly centralized network, information spreads easily but the center is indispensable for the transmission of information.

在本章中，我们将介绍中心性和中心化的概念，这是网络分析中最古老的两个概念。大多数社交网络都包含处于核心地位的人或组织。由于他们的职位，他们可以更好地获取信息和传播信息的更好机会。这被称为以自我为中心的中心性方法。从社会中心的角度来看，整个网络或多或少是中心化的。请注意，我们使用中心性来指代网络中各个顶点的位置，而我们使用中心化来表征整个网络。如果中心和外围之间有明确的边界，则网络是高度集中的。在高度中心化的网络中，信息容易传播，但信息的传递离不开中心。

In this chapter, we discuss several ways of measuring the centrality of vertices and the centralization of networks. We confine our discussion of centrality to undirected networks because we assume that information may be exchanged both ways between people or organizations that are linked by a tie. Concepts related to importance in directed networks, notably prestige, are discussed in Part IV of this book.

在本章中，我们将讨论测量顶点中心性和网络中心化的几种方法。我们将我们对中心性的讨论限制在无向网络上，因为我们假设信息可以在通过纽带连接的人或组织之间以双向方式交换。与有向网络中的重要性相关的概念，尤其是声望，将在本书的第四部分进行讨论。

7 Brokers and Bridges

A person with many friends and acquaintances has better chances of getting help or information. Therefore, social ties are one measure of social capital, an asset that can be used by actors for positive advantage. Network analysts, however, discovered that the kind of tie is important in addition to the sheer number of ties. Their general argument is that strong (i.e., frequent or intense) ties with people who are themselves related yield less useful information than weak ties with people who do not know one another. Having a lot of ties within a group exposes a person to the same information over and over again, whereas ties outside one’s group yield more diverse information that is worth passing on or retaining to make a profit.

有很多朋友和熟人的人更有可能获得帮助或信息。因此，社会联系是社会资本的一种衡量标准，一种可以被行动者用来获得积极优势的资产。然而，网络分析师发现，除了绝对数量的关系之外，关系的种类也很重要。他们的一般论点是，与与彼此不认识的人的弱联系相比，与自己相关的人的强（即频繁或强烈）联系产生的有用信息更少。在一个群体内有很多关系会让一个人一遍又一遍地接触相同的信息，而在一个群体之外的关系会产生更多样化的信息，这些信息值得传递或保留以获利。

As a consequence, we have to pay attention to the ties between a person’s contacts. A person who is connected to people who are themselves not directly connected has opportunities to mediate between them and profit from his or her mediation. The ties of this person bridge the structural holes between others. It is hypothesized that people and organizations that bridge structural holes between others have more control and perform better.

因此，我们必须注意一个人的联系人之间的联系。与自己没有直接联系的人有联系的人有机会在他们之间进行调解并从他或她的调解中获利。这个人的关系弥合了其他人之间的结构漏洞。据推测，在其他人之间架起结构漏洞的人和组织拥有更多的控制权并表现得更好。

In this chapter, we first discuss bridges at the level of the entire network (Section 7.3). Which ties (bridges) and which vertices (cut-vertices) are indispensable for the network to remain connected? If a network contains such ties and vertices, it contains bottlenecks and the flow of information through the network is vulnerable. In the remaining sections, we focus on brokerage at the level of individuals. Who is in the best position to profit from his or her social ties (Section 7.4), and how is this affected by group membership (Section 7.5)?

在本章中，我们首先讨论整个网络级别的网桥（第 7.3 节）。哪些关系（桥）和哪些顶点（切割顶点）对于网络保持连接是必不可少的？如果网络包含这样的关系和顶点，它就会包含瓶颈，并且通过网络的信息流很容易受到攻击。在其余部分中，我们专注于个人层面的经纪业务。谁最能从他或她的社会关系中获利（第 7.4 节），以及团体成员身份如何影响这一点（第 7.5 节）？

8 Diffusion

Diffusion is an important social process. Administrators are interested in the diffusion of information and opinions, manufacturers seek the adoption of new techniques and products, and all of us have a vivid interest in not acquiring contagious diseases. Diffusion processes are being studied in the communication sciences, social psychology and sociology, public administration, marketing, and epidemiology.

扩散是一个重要的社会过程。管理人员对信息和意见的传播感兴趣，制造商寻求采用新技术和新产品，我们所有人都对不感染传染病有着浓厚的兴趣。传播科学、社会心理学和社会学、公共管理、市场营销和流行病学正在研究传播过程。

In this chapter, we present diffusion processes from a network point of view. Diffusion is a special case of brokerage, namely, brokerage with a time dimension. Something – a disease, product, opinion, or attitude – is handed over from one person to another in the course of time. We assume that social relations are instrumental to the diffusion process: They are channels of social contagion and persuasion.

在本章中，我们从网络的角度介绍扩散过程。扩散是经纪的一个特例，即具有时间维度的经纪。随着时间的推移，某种东西——一种疾病、产品、观点或态度——会从一个人传递给另一个人。我们假设社会关系有助于传播过程：它们是社会传染和说服的渠道。

If personal contacts are important, then the structure of personal ties is relevant to the diffusion process and not just the personal characteristics that make one person more open to innovations than another. We inves- tigate the relation between structural positions of actors and the moment at which they adopt an innovation.

如果个人接触很重要，那么个人关系的结构与传播过程有关，而不仅仅是使一个人比另一个人更愿意接受创新的个人特征。我们研究了参与者的结构位置与他们采用创新的时刻之间的关系。

Part IV Ranking

Previous chapters paid little attention to the direction of social relations. In matters of cohesion or brokerage, it is more important to know that a tie exists than to know who initiates it. In this part, however, direction is central, especially asymmetry in social relations. Which choices are not reciprocated? Asymmetry in social relations points to social prestige and ranking.

前几章很少关注社会关系的方向。 在凝聚力或中介问题上，知道存在联系比知道是谁发起的更重要。 然而，在这部分，方向是核心，尤其是社会关系的不对称。 哪些选择不会得到回报？ 社会关系的不对称指向社会声望和等级。

Chapter 9 presents the concept of structural prestige as the reception of direct and indirect positive choices. We compare structural prestige scores with social prestige of persons measured separately to find that there is some but not perfect overlap. In Chapter 10, we discuss techniques to infer informal or latent ranking from the occurrence of different types of triads and acyclic components in a directed network. Finally, Chapter 11 focuses on networks with arcs indicating the passing of time: genealogical networks and citation networks.

第 9 章提出了结构声望的概念，即对直接和间接积极选择的接受。 我们将结构声望分数与单独测量的人的社会声望进行比较，发现存在一些但不完全重叠的重叠。 在第 10 章中，我们讨论了从有向网络中不同类型的三元组和非循环组件的出现来推断非正式或潜在排名的技术。 最后，第 11 章重点介绍带有指示时间流逝的弧线的网络：系谱网络和引文网络。

9 Prestige

In directed networks, people who receive many positive choices are considered to be prestigious. Prestige becomes salient especially if positive choices are not reciprocated, for instance, if everybody likes to play with the most popular girl or boy in a group but he or she does not play with all of them or, in the case of sentiments, if people tend to express positive sentiments toward prestigious persons but receive negative sentiments in return. In these cases, social prestige is connected to social power and the privilege of not having to reciprocate choices.

在有向网络中，接受许多积极选择的人被认为是有声望的。声望变得显着，尤其是在积极的选择没有得到回报的情况下，例如，如果每个人都喜欢和一群中最受欢迎的女孩或男孩一起玩，但他或她并没有和所有人一起玩，或者在感情方面，如果人们倾向于对有声望的人表达积极的情绪，但会收到消极的情绪作为回报。在这些情况下，社会声望与社会权力和不必做出回报选择的特权有关。

In social network analysis, prestige is conceptualized as a particular pattern of social ties. We discuss techniques to calculate the structural prestige of a person from his or her social ties, notably sociometric choices. We do not compute a prestige score for an entire network.

在社会网络分析中，声望被概念化为一种特殊的社会关系模式。我们讨论了从一个人的社会关系，特别是社会计量选择来计算他或她的结构声望的技术。我们不计算整个网络的声望分数。

Structural prestige is not identical to the concept of social prestige in the social sciences or in ordinary speech. For example, the medical profession is thought to be prestigious, but it is difficult to consider professions as a network in which many arcs point toward the medical profession. The prestige of an art museum may depend on the value and origins of its collection rather than on the number of art works it attracts (receives) from other museums. However, social prestige is probably related to structural prestige. In community studies, for example, a physician is more often nominated in advice-seeking relations than members of many other professions, and a prestigious art museum receives more attention from art critics than less prestigious ones.

结构性声望与社会科学或日常言语中的社会声望概念不同。例如，医学专业被认为是有声望的，但很难将专业视为一个网络，其中许多弧线指向医学专业。艺术博物馆的声望可能取决于其收藏的价值和来源，而不是它从其他博物馆吸引（接收）的艺术品数量。然而，社会声望可能与结构声望有关。例如，在社区研究中，医生在寻求建议的关系中比许多其他职业的成员更常被提名，而一个有声望的艺术博物馆比不那么有声望的艺术博物馆受到更多艺术评论家的关注。

In this chapter, we compare the structural prestige of families within a network of visiting ties to their social prestige. As is shown, the two kinds of prestige are related but far from identical. Therefore, be careful not to equate structural prestige to social prestige. Instead, find out whether structural prestige scores on a social relation match indicators of social prestige that are measured by external variables. In a particular setting, which social relation is connected to social prestige?

在本章中，我们将访问关系网络中家庭的结构声望与其社会声望进行比较。如图所示，这两种声望是相关的，但远非相同。因此，请注意不要将结构声望等同于社会声望。相反，找出社会关系的结构声望分数是否与由外部变量衡量的社会声望指标相匹配。在特定环境中，哪种社会关系与社会声望有关？

10 Ranking

In the social sciences, society is regarded as a set of social layers or strata. Instead of ranking people, groups, or organizations on a continuous scale of prestige, they are usually classified into a limited set of discrete ranks, for instance, working class, lower middle class, upper middle class, and upper class. Within a group of humans, discrete ranking also occurs, for instance, leaders, followers, and outcasts. The stratification of art worlds into stars, settled artists, and mediocre artists is likely another example. In this chapter, we discuss techniques to extract discrete ranks from social relations.

在社会科学中，社会被视为一组社会层次或阶层。他们通常被分为一组有限的离散等级，例如工人阶级、中下阶级、中上阶级和上层阶级，而不是按照连续的声望等级对人、团体或组织进行排名。在一群人中，也会出现离散的排名，例如领导者、追随者和被排斥者。将艺术世界分为明星、定居艺术家和平庸艺术家可能是另一个例子。在本章中，我们讨论从社会关系中提取离散等级的技术。

Social ranking may be formal or informal, and the two types of ranking may coexist. In a formal ranking, it is written down who commands whom, and insignia or symbols minimize the ambiguity of the ranking and preclude any confusion about a person’s rank. The army is an obvious example with its elaborate hierarchy. In contrast, an informal ranking is neither written down nor expressed by official symbols. It manifests itself in the opinions and behavior of people toward one another: respect and acts of deference versus disrespect and dominance.

社会排名可以是正式的，也可以是非正式的，两种排名可能并存。在正式的排名中，会写下谁指挥谁，以及标志或符号，最大限度地减少排名的歧义，并避免对一个人的排名产生任何混淆。军队是一个明显的例子，其精心设计的等级制度。相比之下，非正式排名既不写下来也不用官方符号表示。它体现在人们对彼此的看法和行为上：尊重和顺从与不尊重和支配。

The creation and maintenance of an informal ranking is a very important social process. Social network analysis is needed to investigate it and to assess the positions that individuals occupy within the informal ranking. If a formal ranking exists, it is interesting to compare it to the informal ranking because they do not need to match, just like informal communication patterns often deviate from the official communication structure.

非正式排名的创建和维护是一个非常重要的社会过程。需要进行社交网络分析来对其进行调查并评估个人在非正式排名中占据的位置。如果存在正式排名，将其与非正式排名进行比较是很有趣的，因为它们不需要匹配，就像非正式沟通模式经常偏离官方沟通结构一样。

The structural concept of social ranking is an extension to balance theory presented in Chapter 4. Balance theory assigns people to clusters that are not ranked with respect to one another. Within a cluster, people tend to like each other but do not like members of other clusters. Within clusters as well as between clusters, ties are supposed to be symmetric: You are supposed to reciprocate the sentiment or choices that you receive. Elaborating on this perspective, asymmetric ties, for instance, A reports to B but B does not report to A, indicate ranking: B is ranked over A.

社会排名的结构概念是对第 4 章中介绍的平衡理论的扩展。平衡理论将人们分配到彼此没有排名的集群中。在一个集群内，人们往往喜欢彼此，但不喜欢其他集群的成员。在集群内部以及集群之间，关系应该是对称的：你应该回报你收到的情绪或选择。从这个角度详细阐述，不对称关系，例如，A 向 B 报告但 B 不向 A 报告，表明排名：B 的排名高于 A。

11 Genealogies and Citations

Time is responsible for a special kind of asymmetry in social relations, because it orders events and generations in an irreversible way. Social identity and position is partially founded on common ancestors, whether in a biological sense (birth) or in an intellectual manner: citations by scientists or references to predecessors by artists. This is social cohesion by common descent, which is slightly different from cohesion by direct ties (see Part II). Social communities and intellectual traditions can be defined by a common set of ancestors, by structural relinking (families that intermarry repeatedly), or by long-lasting cocitation of papers.

时间是造成社会关系中一种特殊的不对称的原因，因为它以不可逆转的方式对事件和世代进行排序。 社会身份和地位部分建立在共同祖先的基础上，无论是在生物学意义上（出生）还是在智力方面：科学家的引用或艺术家对前辈的引用。 这是共同血统的社会凝聚力，与直接关系的凝聚力略有不同（见第二部分）。 社会社区和知识传统可以由一组共同的祖先、通过结构重新链接（反复通婚的家庭）或通过长期的论文引用来定义。

Pedigree is also important for the retrospective attribution of prestige to ancestors. For example, in citation analysis the number of descendants (citations) is used to assign importance and influence to precursors. Genealogy is the basic frame of reference here, so we discuss the analysis of genealogies first.

血统对于将声望归于祖先也很重要。 例如，在引文分析中，后代（引文）的数量用于为前体分配重要性和影响。 家谱是这里的基本参考框架，所以我们先讨论家谱分析。

Part V Modeling

In the final part, we shift our attention from purely exploratory description of network structure to modeling. Two types of modeling are presented: blockmodels (Chapter 12) and random graph models (Chapter 13).

在最后一部分，我们将注意力从对网络结构的纯粹探索性描述转移到建模上。介绍了两种类型的建模：块模型（第 12 章）和随机图模型（第 13 章）。

Cohesion, brokerage, and ranking are connected to social roles: being a member of a group, being a mediator, or being a superior. Each of these roles is associated with a particular pattern of ties. A blockmodel describes the social roles and associated patterns of ties in the network at large. Blockmodels offer a different perspective on the concepts discussed in previous chapters.

凝聚力、经纪人和排名与社会角色有关：成为团体的成员、调解人或上级。这些角色中的每一个都与特定的联系模式相关联。块模型描述了整个网络中的社会角色和关联模式。 Blockmodels 为前面章节中讨论的概念提供了不同的视角。

Could the structure of our network be random? Network exploration yields values for structural properties of networks, such as cohesion, centrality, and ranking. We assume that the structural properties result from choices made by the people represented in the network – whom to relate to, whom to avoid – and that structural properties are consequential to their attitudes and behavior. The social causes and consequences of network structure give meaning to structural indices. Random graph models show us how confident we can be that network properties and effects result from social rather than random processes.

我们的网络结构可能是随机的吗？网络探索产生网络结构属性的值，例如内聚性、中心性和排名。我们假设结构特性是由网络中代表的人做出的选择产生的——与谁联系，避免谁——并且结构特性对他们的态度和行为产生影响。网络结构的社会因果关系赋予结构指数以意义。随机图模型向我们展示了我们对网络属性和影响来自社会而非随机过程的信心。

12 Blockmodels

In previous parts of this book, we have presented a wide range of techniques for analyzing social networks. We have discovered that one structural concept can often be measured in several ways (e.g., centrality). We have not encountered the reverse, that is, a single technique that is able to detect different kinds of structures (e.g., cohesion and centrality). In this final chapter, we present such a technique, which is called blockmodeling.

在本书的前几部分中，我们已经介绍了多种分析社交网络的技术。我们发现一个结构概念通常可以用多种方式来衡量（例如，中心性）。我们还没有遇到相反的情况，即能够检测不同类型结构（例如内聚性和中心性）的单一技术。在最后一章中，我们将介绍一种称为块建模的技术。

Blockmodeling is a flexible method for analyzing social networks. Several network concepts are sensitive to exceptions; for instance, a single arc may turn a ranking into a rankless cluster (Chapter 10). Empirical data are seldom perfect, so we need a tool for checking the structural features of a social network that allows for exceptions or error. Blockmodeling and hierarchical clustering, which are closely related, are such tools.

块建模是一种用于分析社交网络的灵活方法。一些网络概念对异常很敏感；例如，单个弧可以将排名变成无秩集群（第 10 章）。经验数据很少是完美的，因此我们需要一种工具来检查允许异常或错误的社交网络的结构特征。密切相关的块建模和层次聚类就是这样的工具。

Although blockmodeling is a technique capable of detecting cohesion, core-periphery structures, and ranking, it does not replace the techniques presented in previous chapters. At present, blockmodeling is feasible and effective only for small dense networks, whereas the other techniques work better on large or sparse networks. In addition, blockmodeling is grounded on different structural concepts: equivalence and positions, which are related to the theoretical concepts of social role and role sets. Blockmodels group vertices into clusters and determine the relations between these clusters (e.g., one cluster is the center and another is the periphery). In contrast, the techniques discussed in previous chapters, such as the measures of centrality, compute the structural position of each vertex individually.

尽管块建模是一种能够检测内聚性、核心外围结构和排序的技术，但它并不能取代前几章中介绍的技术。目前，块建模仅对小型密集网络可行且有效，而其他技术在大型或稀疏网络上效果更好。此外，块建模基于不同的结构概念：等价和位置，这与社会角色和角色集的理论概念有关。 Blockmodels 将顶点分组并确定这些集群之间的关系（例如，一个集群是中心，另一个是外围）。相比之下，前几章讨论的技术（例如中心性度量）单独计算每个顶点的结构位置。

Blockmodeling uses matrices as computational tools and for the visualization of results. Therefore, we introduce the matrix as a means for representing social networks before we will proceed to the concept of equivalence and the technique of blockmodeling.

Blockmodeling 使用矩阵作为计算工具和结果的可视化。因此，在我们将介绍等价的概念和块建模技术之前，我们引入矩阵作为表示社交网络的手段。

13 Random Graph Models

The main purpose of social network analysis is detecting and interpreting patterns of social ties among actors (Chapter 1). A pattern of social ties is meaningful if it expresses choices by social actors or the impact of the social system on actors’ behavior and attitudes. Until now, we have implicitly assumed that the observed network expresses choices or social constraint, although we have pointed out that our behavioral interpretations should be checked by comparing them with other indicators – for example, see the discussion on structural and social prestige in Chapter 9.

社交网络分析的主要目的是检测和解释参与者之间的社会联系模式（第 1 章）。 如果一种社会联系模式表达了社会行动者的选择或社会制度对行动者行为和态度的影响，那么它就是有意义的。 到目前为止，我们一直隐含地假设观察到的网络表达了选择或社会约束，尽管我们已经指出应该通过将它们与其他指标进行比较来检查我们的行为解释——例如，参见第 9 章中关于结构和社会声望的讨论 .

In the current chapter, we accept the idea that at least part of the structure of the observed network is random. As a consequence, we should not assume that every pattern found in a network is meaningful. Statistical inference should tell us whether a network characteristic is random. We do not use statistical inference in the classic sense, assuming that the observed network is a random sample from a larger network (design-based inference). For some basic network properties, statistical inference based on a random sample is possible, but that is not what we pursue here. Instead, we present statistical network models that tell us which network characteristics to expect if the lines are assigned to pairs of vertices according to a random process (model-based inference). This approach assumes that network structure could have been different; for example, the line between actors v and u (Figure 128, network C) could have been replaced by a line between v and w (Figure 128, network D), but not every network structure is necessarily equally probable.

在本章中，我们接受观察网络结构的至少一部分是随机的。因此，我们不应该假设在网络中发现的每个模式都是有意义的。统计推断应该告诉我们一个网络特征是否是随机的。我们不使用经典意义上的统计推断，假设观察到的网络是来自更大网络的随机样本（基于设计的推断）。对于一些基本的网络属性，基于随机样本的统计推断是可能的，但这不是我们在这里所追求的。相反，我们提供了统计网络模型，如果根据随机过程（基于模型的推理）将线分配给顶点对，则这些模型会告诉我们预期的网络特征。这种方法假设网络结构可能不同；例如，参与者 v 和 u 之间的线（图 128，网络 C）可以被 v 和 w 之间的线替代（图 128，网络 D），但并非每个网络结构都必然具有同等可能性。

Suppose that we have a transitivity hypothesis for the friendship relation: People are more likely to make friends with their friends’ friends than with other people. If person v is a friend of person z and the latter is befriended by persons u and w, then v is more likely to be a friend of u and w than a person who is not befriended by person z. In social theory, we do not usually formulate deterministic hypotheses – for example, we do not think it is realistic to assume that everyone is a friend of all his/her friends’ friends. If we would, all people linked by paths of friendship ties would also be direct friends, which would raise a person’s number of friends to an incredibly high value for a larger network.

假设我们对友谊关系有一个传递性假设：与其他人相比，人们更有可能与朋友的朋友交朋友。 如果人 v 是人 z 的朋友，而后者与人 u 和 w 成为朋友，那么与没有与人 z 成为朋友的人相比，v 更有可能成为 u 和 w 的朋友。 在社会理论中，我们通常不会制定确定性假设——例如，我们认为假设每个人都是他/她朋友的朋友的朋友是不现实的。 如果我们愿意，所有通过友谊关系路径联系起来的人也将成为直接朋友，这将使一个人的朋友数量增加到一个更大的网络的难以置信的高值。

We prefer to think of tendencies or probabilities. If we encounter a network like C in Figure 128, perhaps v has not met w yet or v currently thinks s/he has enough friends. There can be many different and ad hoc reasons that the observed network contains a line between v and u but not between v and w. We do not care about these ad hoc reasons because we consider their effects random or noise. We are interested in systematic effects on line formation, such as transitivity, which could also have produced a different network, for example, containing (also) a line between v and w or between u and w.

我们更喜欢考虑趋势或概率。 如果我们在图 128 中遇到像 C 这样的网络，可能 v 还没有遇到 w 或者 v 目前认为他/她有足够的朋友。 观察到的网络在 v 和 u 之间包含一条线，但在 v 和 w 之间不包含一条线，可能有许多不同的临时原因。 我们不关心这些临时原因，因为我们认为它们的影响是随机的或噪声。 我们对线形成的系统影响感兴趣，例如传递性，这也可能产生不同的网络，例如，包含（也）在 v 和 w 之间或 u 和 w 之间的线。

As a consequence of randomness, many networks are possible and could have been observed. In principle but usually not in practice, we could list all networks that are possible provided that we formulate conditions such as the number of vertices in the network. Figure 128 shows some of the simple undirected networks with four vertices that could have occurred, ranging from the empty network (A) to the complete network (E). Note that networks containing only intransitive triples may also occur (e.g., Network B in Figure 128), but if our transitivity hypothesis is true, they should have a lower probability to appear.

由于随机性，许多网络都是可能的，并且可以被观察到。 原则上但通常不是在实践中，我们可以列出所有可能的网络，前提是我们制定了网络中的顶点数量等条件。 图 128 显示了一些可能出现的具有四个顶点的简单无向网络，范围从空网络 (A) 到完整网络 (E)。 请注意，仅包含不传递三元组的网络也可能出现（例如，图 128 中的网络 B），但如果我们的传递性假设为真，则它们出现的概率应该较低。

In addition to the set of all possible networks, then, we need the probabilities for each of these networks to occur. The probabilities associated with all possible networks constitute a probability distribution. The probability distributions that we will encounter belong to families, that is, sets of probability distributions that resemble one another. Within each family, one probability distribution differs from the other on one or more characteristics, which are called parameters, such as the average probability of a line to occur within a pair.

除了所有可能网络的集合之外，我们还需要这些网络中的每一个发生的概率。 与所有可能的网络相关联的概率构成一个概率分布。 我们将遇到的概率分布属于家族，即一组彼此相似的概率分布。 在每个系列中，一个概率分布在一个或多个特征上与另一个不同，这些特征称为参数，例如一条线在一对中出现的平均概率。

In the transitivity example, we might assume that each line has the same baseline probability to occur, but every transitive triple created by a line, for example, the transitive triple v–z–u, raises the probability by a fixed amount. In this case, we have a probability distribution with two parameters: a general baseline probability and a transitivity bonus probability on top of that. If we define the collection of possible networks and an accompanying probability distribution in a mathematical way, we have a statistical model for the network, which describes the random process that we assume to have shaped the network.

在传递性示例中，我们可能假设每条线都有相同的基线发生概率，但是由一条线创建的每个传递三元组，例如传递三元组 v-z-u，都会将概率提高固定数量。 在这种情况下，我们有一个带有两个参数的概率分布：一般基线概率和在此之上的传递性奖励概率。 如果我们以数学方式定义可能网络的集合和伴随的概率分布，我们就有了网络的统计模型，它描述了我们假设塑造网络的随机过程。

A statistical network model is a mathematical description of a collection of possible networks and a probability distribution on this set.

统计网络模型是对可能网络集合和该集合上的概率分布的数学描述。

In this chapter, we present statistical models for overall network structure. This approach is mainly interested in global characteristics of networks and it specifies random processes for generating networks with typical structural features such as a particular degree of clustering or a characteristic average path distance. Section 13.3 presents the most popular network models: classic Bernoulli and conditional uniform models, small-world models, and preferential attachment or power-law, scale-free models. We will use these models to construct confidence intervals for network properties in Section 13.4.

在本章中，我们介绍了整体网络结构的统计模型。 这种方法主要对网络的全局特征感兴趣，它指定了生成具有典型结构特征的网络的随机过程，例如特定的聚类程度或特征平均路径距离。 13.3 节介绍了最流行的网络模型：经典的伯努利和条件均匀模型、小世界模型、优先依附或幂律、无标度模型。 我们将在 13.4 节中使用这些模型来构建网络属性的置信区间。

There is another kind of statistical network model focusing on local network structure rather than overall network structure. These models test hypotheses on tie formation: how do actors adapt their lines to other lines in which they or their peers are involved? The models, notably exponential random graph models (ERGM) for cross-sectional data and continuous-time Markov process models for panel data (e.g., as implemented in SIENA software), offer statistical tests on network data that are comparable to hypothesis tests on attribute data that are the mainstay of empirical research in the social sciences. Their discussion, however, deserves a separate book and their application requires other software than Pajek; we refer the reader to the Further Reading section.

还有一种统计网络模型侧重于局部网络结构而不是整体网络结构。 这些模型检验了关于关系形成的假设：演员如何将他们的台词适应他们或他们的同龄人参与的其他台词？ 这些模型，特别是用于横截面数据的指数随机图模型 (ERGM) 和用于面板数据的连续时间马尔可夫过程模型（例如，在 SIENA 软件中实现），提供了网络数据的统计检验，与属性的假设检验相当 数据是社会科学实证研究的主要支柱。 然而，他们的讨论值得单独写一本书，他们的应用程序需要除 Pajek 之外的其他软件； 我们建议读者参阅进一步阅读部分。