SNA AND THE APLP PRESENTATION - East-West … · SNA AND THE APLP 17 Jan 2006 ... Broker is a...

77
SNA AND THE APLP PRESENTATION January 19, 2006

Transcript of SNA AND THE APLP PRESENTATION - East-West … · SNA AND THE APLP 17 Jan 2006 ... Broker is a...

SNA AND THE APLP

PRESENTATION

January 19, 2006

SNA AND THE APLP 17 Jan 2006 v3.ppt - 1 -

OK, SO WHO HAS ONE OF THESE?

OR

SNA AND THE APLP 17 Jan 2006 v3.ppt - 2 -

WHAT IS NETWORK THEORY?

• Focus is on relationships among actors as an explanation of actor and group outcomes

- rather than attributes of actors

Most of the time, both attributes and relationships are needed to explain outcomes

Most of the time, both attributes and relationships are needed to explain outcomes

Source: Barry Wellmann

SNA AND THE APLP 17 Jan 2006 v3.ppt - 3 -

WHY DO WE CARE?

Social Capital� Group level concept: cohesion

• pattern of ties among members of a group confers competitive advantage

- immigrant groups, organizations, countries- Eg. Help with the rice harvest; which village is more likely to

survive?

� Individual level concept: centrality• benefits of being well connected

- material aid- information (broadly defined)- fun, companionship, emotional support, love ...

SNA AND THE APLP 17 Jan 2006 v3.ppt - 4 -

TODAY�S AGENDA

Your egonets

The 2006 APLP network

The APLP Network: The Results (So Far)!• A de-brief of the survey process• What we�ve learned so far

Workshop: Understanding and Building the APLP

SNA AND THE APLP 17 Jan 2006 v3.ppt - 5 -

EGONETSWhat Can You Learn As Individuals About Your Own Networks?

Degree = 14Density = 37.36%Average distance = 1.71Broker = 31%(1)

(1) Broker is a measure of the number of pairs in the ego which are not connected, divided by the number of total pairs in the ego network. It indicates the opportunity for a person to ‘broker’ the creation of links within their own network.

SNA AND THE APLP 17 Jan 2006 v3.ppt - 6 -

DECIPHERING YOUR OWN EGONETS

Country codes

SNA AND THE APLP 17 Jan 2006 v3.ppt - 7 -

SO HOW DO YOU FIT INTO THE OVERALL 2006 COHORT NETWORK?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 8 -

TODAY�S AGENDA

Your egonets

The 2006 APLP network

The APLP Network: The Results (So Far)!• A de-brief of the survey process• What we�ve learned so far

Workshop: Understanding and Building the APLP

SNA AND THE APLP 17 Jan 2006 v3.ppt - 9 -

HOW THINGS LOOKED IN AUGUST 2003Based On Interactions Over First 3 Weeks

SNA AND THE APLP 17 Jan 2006 v3.ppt - 10 -

HOW DO YOU THINK THE NETWORK WILL EVOLVE OVER TIME?

What would you expect to see if we did this exercise again in 3 months? 6 months? 1 year?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 11 -

HOW DO YOU THINK THE NETWORK WILL EVOLVE OVER TIME?

Then Now

?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 12 -

SOME METRICS...

Density � 0.2249

Average path length � 2.365

Clustering Coefficient � 0.425

Cohesion � 0.341

What�s going on?

Density � .4052

Average path length � 2.564

Clustering Coefficient � 0.812

Cohesion � 0.224(range 0 to 1; larger values

indicate greater cohesiveness)

Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 13 -

THE 2006 COHORT NETWORK SEVERAL MONTHS LATER

20022003200420052006Staff

n=68

SNA AND THE APLP 17 Jan 2006 v3.ppt - 14 -

POSITION OF APLP STAFF IN OVERALL NETWORK

Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 15 -

WHAT HAPPENS IF WE TAKE THE STAFF OUT?

Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 16 -

NETWORK STRUCTURES DIFFER DRAMATICALLY DEPENDING ON TYPE OF LINK

1=Social, 2=Collaboration, 3=Advice/Support, 4=Information exchange

1 23 4

SNA AND THE APLP 17 Jan 2006 v3.ppt - 17 -

NETWORK STRUCTURES DIFFER DRAMATICALLY DEPENDING ON TYPE OF LINK

1=Social

1Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 18 -

NETWORK STRUCTURES DIFFER DRAMATICALLY DEPENDING ON TYPE OF LINK

2=Collaboration

2Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 19 -

NETWORK STRUCTURES DIFFER DRAMATICALLY DEPENDING ON TYPE OF LINK

3=Advice/Support

3

No staff included!

Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 20 -

NETWORK STRUCTURES DIFFER DRAMATICALLY DEPENDING ON TYPE OF LINK

4=Information exchange

4

Then Now

SNA AND THE APLP 17 Jan 2006 v3.ppt - 21 -

WHO ARE THE HUBS? HAS THIS CHANGED?Degree Centrality

SNA AND THE APLP 17 Jan 2006 v3.ppt - 22 -

WHO HAS CONTROL OVER INFORMATION? HAS THIS CHANGED?

Between-ness Centrality

SNA AND THE APLP 17 Jan 2006 v3.ppt - 23 -

HOW HAS THE MOST INFLUENCE? HAS THIS CHANGED?Eigenvector Centrality

SNA AND THE APLP 17 Jan 2006 v3.ppt - 24 -

TODAY�S AGENDA

Your egonets

The 2006 APLP network

The APLP Network: The Results (So Far)!• A de-brief of the survey process• What we�ve learned so far

Workshop: Understanding and Building the APLP

SNA AND THE APLP 17 Jan 2006 v3.ppt - 25 -

THE SNA TEAM!

SNA AND THE APLP 17 Jan 2006 v3.ppt - 26 -

DETAILED WORKPLAN (PROPOSED)September

1 2 3 4

October

1 2 3 4

November

1 2 3 4

December

1 2 3 4

TimingEvent

Finalize research design• Objectives and hypotheses• Team members roles• Data collection methods• Detailed work plan

Data collection instrument• Develop instrument• Test with focus group• Refine and finalize

Develop strategy to reach out to alumni and elicit participation

Administer survey

Conduct focus groups?

Compile and clean data

Data analysis

Write up findings

Present findings

Meetings and weekly calls

SNA AND THE APLP 17 Jan 2006 v3.ppt - 27 -

DESIGNING AND FINALIZING THE SURVEY INSTRUMENT

Hypothesis drivenTestedMany iterations � �beat it up�Strengths and challenges

SNA AND THE APLP 17 Jan 2006 v3.ppt - 28 -

SURVEY RESPONSES109 responses (out of 202 possible) - 54%!

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

2002 2003 2004 2005 2006

% of population responding

By Year

0

10

20

30

40

50

60

70

80

Men Women

By Gender# of surveys received

9 15 22 26 37How does this compare with the overall

gender ratio of the program?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 29 -

GETTING PEOPLE TO RESPONDPleading And Cajoling Went So Far; Bribery Also Worked

0

10

20

30

40

50

60

70

80

90

100

30-Oct

3-Nov

7-Nov

11-Nov

15-Nov

19-Nov

23-Nov

27-Nov

1-Dec

5-Dec

9-Dec

13-Dec

Number ofresponsesreceived

Surveys first go out by

emal

Nick sends email (Nov 9)

Scott threatens �summative

reviews� for 2006 class

Eric appeals directly to class

reps for help

Nick announces survey prize

draw

Deadline Dec 1

SNA AND THE APLP 17 Jan 2006 v3.ppt - 30 -

SOME OF OUR HYPOTHESES

The existing network is clustered around densely connected cohort groups

Information flows efficiently and effectively within cohort groups, but poorly between cohort groups

New cohort group networks are initially hierarchical and centralized around program staff. The centrality of program staff to the network decreases over time, and the network becomes less clustered around hubs

The density of the cohort networks, and the overall network decreases over time

Dimensions of diversity (gender, age, ethnicity, religion, other?) and/or physical proximity or shared professional interests impact the structure of network and are the leading cause of the development of clusters within cohort groups, as well as the strength of the ties and how long they endure over time

What can we learn about how and why strong networks/relationships form, so that their

development can be nurtured and supported?

What can we learn about how and why strong networks/relationships form, so that their

development can be nurtured and supported?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 31 -

TODAY�S AGENDA

Your egonets

The 2006 APLP network

The APLP Network: The Results (So Far)!• A de-brief of the survey process• What we�ve learned so far

Workshop: Understanding and Building the APLP

SNA AND THE APLP 17 Jan 2006 v3.ppt - 32 -

A FIRST GLIMPSE

What does it mean? Insights? Questions to explore? Hypotheses we should test?

What does it mean? Insights? Questions to explore? Hypotheses we should test?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 33 -

UNDERSTANDING WHAT�S GOING ONSOME FIRST STEPS

What is the question we are trying to answer? What is the nature of the links/ties?

• what is the nature of the tie? (social, advice, etc) • how strong are the relationships?• who is more strongly connected to whom?

What are the node attributes?• who is in/out of the network (eg. staff vs non staff, cohort year, etc)• color/size/shape the nodes to reflect country of origin, cohort year,

gender, staff vs non staff, country of residence, etc. ?• does the visualization start to tell a story?

What are the applicable network metrics?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 34 -

STRENGTH OF TIES

SNA AND THE APLP 17 Jan 2006 v3.ppt - 35 -

ONE CLUSTER IS VERY TIGHTLY CONNECTED � FREQUENCY OF CONTACT MORE THAN ONCE PER WEEK

Hypothesis: There is a reason that this cluster

is so much more strongly connected than other clusters. What are

the reasons for this? Homophily, (attributes), shared experiences, use

of technology etc.

SNA AND THE APLP 17 Jan 2006 v3.ppt - 36 -

SOME BASIC NETWORK PARAMETERSWhat kinds of questions do these measure?

Degree (in non directional networks), In degree and out degree (in directional networks)

• Who in the network has the highest number of ties?- What attributes seem to influence number of ties?

• Average degree (�connected-ness�) of survey respondents?Density

• Just how tightly connected is the network?• Number of ties, expressed as percentage of the number of

ordered/unordered pairsGeodesic distance or average path length

• How quickly can you get from one part of the network to another?- Information, contacts, news, gossip...?

Centralization• Does the network revolve around a single node or nodes?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 37 -

WHO HAS THE HIGHEST DEGREE (CONNECTIONS) IN THE NETWORK?

What attributes seem to influence number of ties?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 38 -

DOES GENDER MATTER?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 39 -

DOES COUNTRY OF ORIGIN MATTER?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 40 -

DOES STAFF VERSUS NON STAFF MATTER?

Can we make a conclusion based

solely on the visualization?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 41 -

DOES COHORT YEAR MATTER?

What else does this picture tell you?

20022003200420052006Staff

SNA AND THE APLP 17 Jan 2006 v3.ppt - 42 -

NUMBER OF PEOPLE IN EGO NETWORK From Answering Question On Survey About The Number Of People They Stay In

Touch With

0

5

10

15

20

25

30

35

40

45

50Number of respondents

33 people in contact with more than 20 people more than

once per week? (But did they fill out

information about all these people in their

surveys?)

20+16-2011-156-101-5Not in regular communication

with anyone

SNA AND THE APLP 17 Jan 2006 v3.ppt - 43 -

NUMBER OF PEOPLE IN EGO NETWORK: INCLUDES DEGREE OF ALL NODES

From Network Data; Both In-degree and Out-degree

0

10

20

30

40

50

60

70

80Number of respondents

20+16-2011-156-101-5Not in regular communication

with anyone

What happened to all those

people reporting more than 20

contacts?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 44 -

NUMBER OF PEOPLE IN EGO NETWORK: RESPONDENTS OUT-DEGREE ONLY

From Network Data

0

10

20

30

40

50

60

70

80

90Number of respondents

20+16-2011-156-101-5Not in regular communication

with anyone

Why were people reporting higher

number of contacts but not specifying all of

them in their network data?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 45 -

DENSITY

Density• Just how tightly connected is the network?• Number of ties, expressed as percentage of the number of

ordered/unordered pairs• 0.0994 for APLP network � low density � partly due to missing data?• Somewhat of a relative measure so need to use caution in interpreting

SNA AND THE APLP 17 Jan 2006 v3.ppt - 46 -

GEODESIC DISTANCE OR PATH LENGTH

Geodesic distance or average path length• How quickly can you get from one part of the network to another?

- Information, contacts, news, gossip...?! Path: can�t repeat node, Trail: can�t repeat line, Walk:

unrestricted! Length of a path is number of links! Distance between two nodes is length of shortest path (aka

geodesic)• Average distance (among reachable pairs) = 4.811• Distance-based cohesion = 0.134• (range 0 to 1; larger values indicate greater cohesiveness)• Distance-weighted Fragmentation = 0.866• Somewhat of a relative measure so need to use caution in interpreting

• Are we more or less cohesive than other Alumni networks?

How many links, and where, do we need to add to decrease average path length?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 47 -

CENTRALIZATION

To what extent does the network revolve around a single node or nodes?

Freeman�s between-ness centrality• a measure of information control• measures the % of paths in the network which go through a single node

• 34.79%

SNA AND THE APLP 17 Jan 2006 v3.ppt - 48 -

CONNECTIVITYHow connected is the network overall? How easily does it break apart?

• Components• Maximal sets of nodes in which every node can reach every other by some

path (no matter how long)• � A connected graph has just one component

• Isolates- Nodes not connected to others

• Pendants• Nodes with only one tie to the network

• Cutpoints- Nodes which, if deleted, would disconnect network

• Bridges• A tie that, if removed, would disconnect net

• Cohesion- Line connectivity is the minimum number of lines that must be removed to

disconnect network- Node connectivity κ is minimum number of nodes that must be removed to

disconnect network

SNA AND THE APLP 17 Jan 2006 v3.ppt - 49 -

Can we identify:Components

IsolatesPendantsCutpointsBridges?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 50 -

CUTPOINTS MARKED IN BLUE; REMOVAL OF 16 CUTPOINTSBREAKS THE NETWORK INTO 20 BLOCKS

Why does this seem

counterintuitive?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 51 -

WHEN WE TAKE AWAY THE APLP STAFF, THE NETWORK STARTS TO BREAK APART

What does the �non staff� network look like? What are its

parameters? Should we include Alumni working as staff?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 52 -

NETWORK WITHOUT STAFF, COLORED BY COHORT

20022003200420052006Staff

SNA AND THE APLP 17 Jan 2006 v3.ppt - 53 -

NETWORK WITHOUT STAFF AND 2006 COHORTREFORMATTED

What is the density of this network? Compared

to the full network? What are the

characteristics of the individuals with highest

degree (hubs)? How strong are these links? What do the strongly linked nodes have in

common?

20022003200420052006Staff

SNA AND THE APLP 17 Jan 2006 v3.ppt - 54 -

SO HOW IMPORTANT ARE SUBGROUPS TO THE NETWORK?

We measure this by looking at transitivity and clustering• tendency towards dense local neighborhoods � ie. most of the people

we know also know each other• �Small world� -- a combination of short average path lengths over the

entire graph, coupled with a strong degree of "clique-like" local neighborhoods

For the APLP community• Overall graph clustering coefficient: 0.843• Weighted Overall graph clustering coefficient: 0.597 or 59.7%• Quite high relative to overall density of the network of ~ 10%

We can conclude that clustering (subgroups) is an important aspect of the overall network

We can conclude that clustering (subgroups) is an important aspect of the overall network

SNA AND THE APLP 17 Jan 2006 v3.ppt - 55 -

WHAT DO THESE SUBGROUPS LOOK LIKE?Ucinet Has Several Methods Of Determining Subgroups

Structural definitions of groups � of questionable value!� Clique

• Maximal set of actors in which every actors is connected to every other- � Maximum density (1.0)- � Minimum distances (avg = 1)- � overlapping

� N-clique, n-clan, n-club• A set of nodes that are within distance n of each other• Relaxes distance aspect of clique concept (1-clique is just a clique)

� K-core, K-plexes• A set of n nodes in which every node has a tie to at least n-k others in the set• In a 1-plex, every node is connected to all but one others in the set (i.e., is a clique)

� Factions• A set of mutually exclusive groups of actors such that density of ties within group

is greater than density of ties between groups

What are our hypotheses? How do we test them?What are our hypotheses? How do we test them?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 56 -

7 FACTIONS COMPARED WITH COHORT YEAR

Factions program puts a lot more nodes in the grouping which includes the 2006 class. Why is this? Density of ties within group is greater than

between groups. Cohort an imperfect predictor of subgroups?

Factions program puts a lot more nodes in the grouping which includes the 2006 class. Why is this? Density of ties within group is greater than

between groups. Cohort an imperfect predictor of subgroups?

cohortsfactions

20022003200420052006Staff

SNA AND THE APLP 17 Jan 2006 v3.ppt - 57 -

cohortsK-cores

K-CORES � COMPARE WITH COHORTS?

K-cores puts the �pendants� in a group because it is based on number of ties. Do the K-core results

tell us anything?

K-cores puts the �pendants� in a group because it is based on number of ties. Do the K-core results

tell us anything?

20022003200420052006Staff

SNA AND THE APLP 17 Jan 2006 v3.ppt - 58 -

CENTRALITY, POWER AND INFLUENCE

Degree • the number of nodes adjacent to given node (we already looked at this!)• those with highest in degree are hubs, out degree are �link mavens�

Eigenvector• iterative version of degree centrality• a node�s centrality is proportional to the sum of centralities of those it has ties to• its not who you know, its who they know

Closeness• Sum of geodesic distances to all other nodes• Inverse measure of centrality

Between-ness • number of times that a node lies along the shortest path between two others• measure of control over information flows in a network

SNA AND THE APLP 17 Jan 2006 v3.ppt - 59 -

DEGREE CENTRALITY

SNA AND THE APLP 17 Jan 2006 v3.ppt - 60 -

EIGENVECTOR CENTRALITY

SNA AND THE APLP 17 Jan 2006 v3.ppt - 61 -

BETWEENNESS CENTRALITY

SNA AND THE APLP 17 Jan 2006 v3.ppt - 62 -

DIFFERENT �TYPES� OF NETWORKS

CollaborationSocial

Other

Professional advice

Personal adviceInformation

SNA AND THE APLP 17 Jan 2006 v3.ppt - 63 -

INDIVIDUAL COHORT YEARS

20022003200420052006Staff

2006 2005

2003

2004

2002

SNA AND THE APLP 17 Jan 2006 v3.ppt - 64 -

GEOGRAPHIC DISTRIBUTION OF RESPONDENTS

SNA AND THE APLP 17 Jan 2006 v3.ppt - 65 -

ADDITIONAL (NON NETWORK) FINDINGS

Number of people planning on attending the reunionWho�s willing to help out with whatComments and suggestions

SNA AND THE APLP 17 Jan 2006 v3.ppt - 66 -

ADDITIONAL FINDINGS (NON NETWORK): SUGGESTIONS AND COMMENTS

�good job� rather than job meaning emplyment

This cluster all about requests for regular meetings of alumni

SNA AND THE APLP 17 Jan 2006 v3.ppt - 67 -

SOME TECHNICAL CONSIDERATIONS

Missing data...or the statistical implications of �sampling� in SNA (Basu?)

Directed vs non directed ties

Symmetry and reciprocity (reciprocated ties)• When can we assume this and when cant we?

Valued ties vs non valued ties

Informant accuracy

SNA AND THE APLP 17 Jan 2006 v3.ppt - 68 -

Valued

Not valued

Strength of Tie

NETWORK PARAMETERS DETERMINE THE KIND OF METRICS/ANALYSES WE CAN USE

Where Does The APLP Data Fit?

Directed Non Directed

SNA AND THE APLP 17 Jan 2006 v3.ppt - 69 -

WHERE DOES OUR DATA FIT?

Professional AdvicePersonal Advice

Information

SocialCollaboration

Professional AdvicePersonal Advice

Information

Social Collaboration

Valued

Not valued

Directed Non Directed

Strength of Tie

“Symmetrize”

“Dichotom

ize”

SNA AND THE APLP 17 Jan 2006 v3.ppt - 70 -

DIRECTED TIES

How many of these ties are reciprocated?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 71 -

RECIPROCATED TIES ARE IN REDSMALL PROPORTION OF TOTAL

Percentage of reciprocated ties (out of all existing ties): 13.78%

(#(x->y AND x<-y)/#(x->y OR x<-y)

SNA AND THE APLP 17 Jan 2006 v3.ppt - 72 -

PART OF THE ISSUE IS MISSING DATA?How much of the lack of reciprocity can be accounted for by missing data?

What are some other reasons that might account for

lack of reciprocity?

109 respondents211 nodes (includes staff)

SNA AND THE APLP 17 Jan 2006 v3.ppt - 73 -

TODAY�S AGENDA

Your egonets

The 2006 APLP network

The APLP Network: The Results (So Far)!• A de-brief of the survey process• What we�ve learned so far

Workshop: Understanding and Building the APLP

SNA AND THE APLP 17 Jan 2006 v3.ppt - 74 -

IDEAS FOR SOME �REAL TIME� ANALYSIS

Individual cohorts � What do these networks look like? What are their various SNA metrics? How do they compare? Which cohorts appear to be the most connected? What attributes appear to be driving �connectedness� with the cohort?

Cross cohort analysis � How tightly are the cohort networks connected to each other? What are the characteristics of the �bridges�? How can we strengthen these connections?

Types of networks � Social, advice (personal and professional), information, collaboration. What are their various SNA metrics? Which appear to have the strongest ties? How do they compare? What attributes appear to be driving �connectedness� within the different networks? Which ones do we focus on strengthening and how?

....and so on....!

SNA AND THE APLP 17 Jan 2006 v3.ppt - 75 -

ISSUES FOR DISCUSSION

Have we met our objectives with the SNA survey? Why or why not? What would we do differently next time? (Check out success metrics). What more needs to be done?

Value of the survey and survey data. Should we make this into a longitudinal study? What are some of the challenges of doing a SNA study every year, and how might these be overcome?

What questions are still outstanding? What further analyses or data might help us answer these questions?

Given our findings, what should some general objectives be regarding strengthening the network? For example:

• Cohorts - Do we want to strengthen within cohort networks? If so, what are some ways to do this? What about the larger network as a whole? Strengthen cross-cohort links? How?

• Network type � What types of linkages are of most value? Do we want to build certain types of linkages (eg. professional) that might make the network even more valuable to its members?

• Role of hubs � What are the characteristics of the �hubs�? Is there a way to involve them in strengthening the network (Don�t forget confidentiality issues!) What about reducing dependence on staff as central to the network?

SNA AND THE APLP 17 Jan 2006 v3.ppt - 76 -

Country Total Code USA 71 1 USA Guam 2 1 China 17 2 Vietnam 2 3 Thailand 14 4 Philippines 11 5 Canada 7 6 Bhutan 5 7 Indonesia 5 8 Japan 5 9 Korea 5 10 Myanmar 5 11 Nepal 5 12 Cambodia 4 13 India 4 14 Singapore 4 15 Fed. States of Micronesia 3 16 Tonga 3 17 Uzbekistan 3 18 Hong Kong 2 19 Kyrgyzstan 2 20 Laos PDR 2 21 Mongolia 2 22 Sri Lanka 2 23 Taiwan 2 24 China TAR 1 23Other:

American Samoa 1 26 Belize 1 26 Brazil 1 26 Iran 1 26 Malaysia 1 26 Mexico 1 26 New Zealand 1 26 Pakistan 1 26 Palau 1 26 Papau New Guinea 1 26 Slovakia 1 26 Sweden 1 26 UK 1 26

Network Type CodeProfessional or career advice 1Share information 2Personal advice 3Collaboration 4Social 5Other 6

Rank of Network Type CodeMost frequent reason for communicating 1Second most frequent reason for communicating 2Third most frequent reason for communicating. 3

Respondent CodeNo 0Yes 1

Staff CodeNo 0Yes 1

Year CodeStaff 12002 22003 32004 42005 52006 6

Gender CodeMale 0Female 1