NUTS AND BOLTS:�RESEARCH DESIGN FOR EGOCENTRIC NETWORKS

Brea L. Perry

Duke - Social Networks and Health

ROADMAP

• Deciding what kind of network data you need
• Research design for ego networks
• Software for collecting network data

​

​

​

WHAT KIND OF NETWORK DATA DO YOU NEED?

​

EGOCENTRIC OR SOCIOCENTRIC?

Egocentric

• Measures individuals’ connections to their own personal community networks
• Goal is to gather information about many non-overlapping networks from the standpoint of the social actors situated within them
• Many local networks

​

​

​

​

EGOCENTRIC OR SOCIOCENTRIC?

Sociocentric

• Maps the overall structure of the network, including all direct and indirect ties between actors
• Identifying ties (and the absence of ties) between every member of a bounded (e.g., socially, geographically) group using a census survey
• One global network

​

​

MORE OF A CONTINUUM

Typology of studies based on ego network vs sociocentric data or analysis

​

COOL STUFF ABOUT EGO NETWORKS

Flexibility in data collection

• Because egocentric SNA uses individuals as cases, potential sampling frames and data collection strategies are virtually limitless
• Egocentric data collection tools can easily be incorporated into large-scale or nationally-representative surveys being fielded for a variety of other purposes

​

COOL STUFF ABOUT EGO NETWORKS

Generalizability

• With sociocentric SNA, you observe one set of nodes – one group, one org, one geographic area. Generalization to other groups, even similar ones, is tenuous.
• With ego network designs, you can use any sampling framework, and unconnected (i.e., independent) networks are preferable. You observe many ego networks, allowing generalization to other ego networks collected using the same methods.

​

COOL STUFF ABOUT EGO NETWORKS

Unbounded versus bounded networks

• Sociocentric SNA collects data on ties between all members of a socially or geographically-bounded group and thus has limited inference beyond that group (or other groups like it)
• Egocentric SNA assesses individuals’ personal community networks across any number of social settings, and is therefore less limited in theoretical and substantive scope

​

COOL STUFF ABOUT EGO NETWORKS

Theoretical implications of unboundedness:

• Ability to transcend the boundaries of a single group or domain
• Census in one domain will omit important interaction partners outside that domain
• Makes egocentric ideal for studying what happens to individuals – who operate in multiple contexts in the real world

DISADVANTAGES OF EGOCENTRIC METHODS

• Heavy respondent burden which increases exponentially with network size
• Inability to map the broader social structure in which personal networks are embedded
• Inability to assess the implications for ego of ties that DO NOT exist

GETTING EGOCENTRIC NETWORK DATA

​

GENERATORS

Only ego knows who is in his or her network!

• Generators are questions used to elicit alters
• Followed by questions (name interpreters) about each alter

GENERATORS

Maybe the single most important

decision you will make in designing

egocentric research!

​

Generators establish “operational boundaries on the interpersonal environment (Marsden 1987:123)”

GENERATORS

Elicitation will always be biased because people are not stored randomly in memory

​

• Sources of bias:
• Association (e.g., family first), recency, and typicality
• Respondents have varying levels of energy and interest
• Compromised memories or cognitive deficiencies for some target populations
• Questions that precede generators can impact the way names are recalled

​

NAME GENERATORS

“Boundary definition” strategies:

​

• Cast a wide net
• Capture strong and weak ties of many kinds
• Allows for analytic decisions later on
• Larger networks provide more informative structural measures
• More labor intensive

NAME GENERATORS

McCarty and colleagues (2007):

“You know them and they know you by sight or by name. You have had some contact with them in the past two years, either face-to-face, by phone, mail or e-mail, and you could still contact them if you had to.”

• Strategy elicits large networks containing up to 45 alters
• Minimum of 15 alters is required for computing stable estimates of structural measures like density and centrality

NAME GENERATORS

“Boundary definition” strategies:

​

• Get specific
• Tailor name generator to outcomes of interest
• Functional specificity theory
• More theoretically and substantively driven
• More specific = better reliability (less forgetting)
• Risky, but lower cost resource-wise

NAME GENERATORS

PhenX Toolkit – Health matters name generator

​

• I’m interested in the people in your life that you talk to about health problems when they come up. Who are the people that you discuss your health with or you can really count on when you have physical or emotional problems?

​

• Who are the people, whether or not you have mentioned them before, who are always talking about your mental and physical health and trying to get you to do things about them?

​

• Are there people who are, in general, a burden to you because of their emotional or physical health problems and always want to talk to you about their problems, whether you want them to or not?

​

NAME GENERATORS

NAME GENERATORS

“Boundary definition” strategies:

​

• A hybrid: Multiple name generators
• More valid and reliable data because several types of relationships can be concretely defined, rather than using an abstract criterion that is subject to error
• Methodological safeguard against omitting particular types of ties
• Can examine multiplexity

​

NAME GENERATORS

NAME GENERATORS

Should you ask for a fixed number of alters?

​

• (+) provides an anchor value and increases similarity across networks
• (+) allows for elicitation of weaker ties if number is relatively large (e.g., 25-50)
• (-) may force elicitation of people that don’t matter, or constrain elimination of those that do

​

DO NOT RECOMMEND FOR SMALL NETWORKS!!!

NAMING CONVENTIONS?

Names or initials?

• IRB may not allow you to use full names
• With many alters, ego will need a name that they recognize later in the interview (even MORE important if longitudinal)
• First name and last initial is preferable, but combo WilSha (for William Shakespeare) also works
• Now good software for finding matches if names are not allowed

NAME INTERPRETERS

Common questions about alters

• Demographics (age, sex, race, etc.)
• Strength (closeness, duration, freq of contact)
• How they met (family, work, neighbor, school)

​

NAME INTERPRETERS

Common questions about alters

• Communication (modes, intimacy, trust )
• Support (what they do for R, how much, how well)
• Questions specific to the RQ or outcome
• Do they smoke?
• Are they involved in a social movement?
• Did they help you with your job search?
• Did they think you should seek MH services?

​

NAME INTERPRETERS

Response categories: Nominal (yes/no) or ordinal?

• Yes/no are cognitively easier to determine (more reliable) and faster
• But yes/no provides no discrimination among levels
• Compromise is to use a small Likert scale (e.g., three response categories)

ALTER-ALTER TIES

• To assess structural properties, must ask about relationships between alters
• Typically only ask one alter-alter question because this is very time consuming

​

ALTER-ALTER TIES

Some options:

• “Think about <alter1> and <alter2>. How close are they to one another? (very, sort of, don’t know each other at all)”
• “Think about the relationship between <alter1> and <alter2>. Would you say that they are strangers, acquaintances, or fairly close?”
• “How likely is it that <alter1> and <alter2> talk to each other when you are not around? That is, how likely is it that they have a relationship independent of you? (very, somewhat, not at all)”

​

​

ELECTRONIC DATA COLLECTION �WITH NETWORK CANVAS

​

https://www.networkcanvas.com/

​