Digital Photography

1
Network Cluster Analysis of Departmental
Libraries Used by Faculty at a Large University
Library
JoAnn Jacoby Associate Professor of Library
Administration and Anthropology Sociology
Subject Specialist
7th Northumbria International Conference on
Performance Measurement in Libraries
Information Services
Spiers Estate, South Africa 13-16 August 2007
2
Setting University of Illinois Library

• Major research library big collections (10
  million vols.) great librarians
• Decentralized library system
• (40 libraries)
• Pressures catalyzing change
• User expectations multiplying increasingly
  diverse

3
Purpose

• Map connections between departmental libraries
  from the users perspective
• Gain insight into the use of the 40 libraries by
  faculty across disciplines
• Evaluate how the librarys structure might be
  changed to better serve faculty

4
Methods

• Mined data from survey of all faculty, academic
  professionals staff, 2006
• lthttp//www.library.uiuc.edu/assessment/gt
• Responses to Indicate your primary library/
  (ies) were used to create network maps and
  cluster trees
• Respondents could choose as many libraries as
  they pleased

5
Response Rate
Distributed by a campus-mediated massmail, the
survey was sent to 12,399 (not 13103) individuals
because the headcount at the time was less than
the annual maximum. The response rate was
therefore slightly higher than shown.
6
Social Network Analysis

• Seeks to reveal the patterning of people's
  interaction (Roger Brown 1965)
• Web of social connections key to understanding
  individuals, groups societies
• Studies patterns in the relations social actors
  have with particular others (Grannis 2005)
• Actors can be people, firms, nations, etc.
• Relations whether actors go to war, have sex,
  use a particular library, etc.

7
Methods

• Sociomatrix Data for each relation presented in
  a two-way matrix, where the rows columns refer
  to the actors making up the pairs.
• Format used by most network analysis software
• Adjacency matrix type of sociomatrix used for
  dichotomous data (presence or absence of
  affiliation)

8
Methods
Setting up the matrix
9
Methods

• Two-mode affiliation network
• 1st mode 1350 faculty staff respondents
• 2nd mode 47 libraries
• Connections among members of one of the modes
  are based on linkages established through the
  second mode
• Affiliations and Overlapping Subgroups. In
  Social Network Analysis Methods and
  Applications. Wasserman Faust (1994). Cambridge
  University Press.

10
Methods
Setting up the adjacency matrix- Quantifies the
ties for the relation in question
11
Network map of connections among all libraries.
Lines were drawn when gt5 respondents reported
using both libraries. Selected nodes are shown
in color.
12
Software Packages

• International Network for Social Analysis (INSA),
  Computer Programs for Social Network Analysis
• lthttp//www.insna.org/INSNA/soft_inf.htmlgt
• Comprehensive list of programs for data
  conversion, network analysis, data collection
  project management

13
The INSA website -- a great resource
14
Pajek lthttp//vlado.fmf.uni-lj.si/pub/networks/pa
jek/default.htmgt Free, well-documented, Windows
software
15
TouchGraph lthttp//www.touchgraph.com/gt Commercial
source for software consulting
16
BibApp lthttp//code.google.com/p/bibapp/gt Maps
researchers citations? 'Institutional
Bibliography'
17
Drawn by Hand in Adobe Illustrator (gt20 only
shown here)
18
Size classes by number of connections - Used
natural breaks - Basis for circle size
19
Results

• Network maps were generated showing the
  connections among
• individual faculty maps
• specific departmental libraries
• groupings of related departmental libraries

20
Individual Respondents
21
Pink Cinema Studies, Assistant Professor Indigo
- Physics, Professor Green - Agricultural Science
(ACES), Academic Professional Light Blue -
Spanish, Italian Portuguese, Associate Professor
22
(No Transcript)
23
Specific Libraries
24
City Planning Landscape Architecture Library
25
Modern Languages, English, Classics Latin
American
26
Health Science Library
27
Broad Disciplinary Groupings
28
All Divisions coded by color
29
Area Studies
Arts Humanities
Life Sciences
Physical Sciences Engineering
30
Allied
Central Public Services
Social Sciences
Special Collections
31
Mean Number of Libraries Used by Division
32
Cluster analysis of all libraries
Confirms some patterns in the network maps, but
is of limited utility in identifying potential
mergers
33
Correlations with other indicators of activity
level
34
Relationships between Number of Primary Library
Designations, Circulation Web Transactions
35
So what?

• Faculty dont come in to the library anymore.
• if they can get what they need online
• ..right?

36
Comments Library as Place

• I mostly use the library via online accessIt's
  especially great when working at home or doing
  field research abroad
• Remote delivery of materials is a great
  serviceas my duties make it challenging to get
  onto central campus to find a book.
• Your questions imply that we actually go to the
  library and want to interact with librarians

37
Effect of Online Library Resources by Status
38
Mode of Library Use by Status
39
Comments Library as Place

• Person to person assistance can be key,
  especially when beginning new projects or trying
  to navigate through new search tools.
• It would be easier to find print journals and
  books if the library was open more hours
• The system of departmental libraries makes it
  very difficult for an interdisciplinary
  researcher to find and collect books s/he needs.

40
(No Transcript)
41
Research into Practice

• ongoing decisions regarding the consolidation of
  library service points
• revealing areas of potential collaboration
• e.g., the author collaborated with the Applied
  Health Sciences Library (alx) to create a health
  information portal after seeing evidence of a
  strong shared constituency

42
Conclusions

• Network maps and cluster trees
• reveal connections and patterns in complex
  systems
• provide a data visualization tool that can help
  inform library decision-making

43
Conclusions

• Could be applied to other types of data
• co-locate/bundle suites of services frequently
  used or highly valued by certain groups of
  library users
• create browsing collections (virtual or tangible)
  based on people who checked this out, also
  checked out

44
Questions?