To give you an opportunity to explore social networks through an online application
(http://www.netvis.org)
that integrates data analysis and visualization. The online application is freely available,
and was built with all open-source code
(see http://www.netvis.org/code.php).
Additional resources for understanding social network analysis and visualization
are also available on the web site (see
http://www.netvis.org/resources.php).
You are encouraged to play with the demo before importing your data (see
http://www.netvis.org/demo.php,
and click on the "Test Data" link).
Research Problem
The most common social science approach
to data analysis and visualization involves variables (often columns in a spreadsheet)
and cases (often rows in a spreadsheet). Variables are measured constructs,
like age, size, or effectiveness. Cases are part of the sample, such as individuals,
groups, or organizations. Using this approach, then, you could compute as well
as graph the average age, size, or effectiveness of the individuals, groups,
or organizations. However, what if you wanted to know whether older individuals
were more likely than younger individuals to communicate with other older individuals?
Or whether larger groups tended to have sparse rather than dense communication
patterns compared with smaller groups? Or whether effective organizations had
less core-periphery communication structure than ineffective organizations?
You could only answer these questions if you collected social network data,
that is, data reflecting relations or connections. In addition to variables
and cases, you need matrices.
Tutorial
Login to this web page above
(or see
http://www.netvis.org/register.php if you haven't registered yet), and
be sure to SAVE your answer for each question. The tutorial should take
approximately two hours, and at a minimum, include two survey responses
(one person from each social network you select for your study)
(Feel free to complete this tutorial within Microsoft Word first
so that you have a version saved on your harddrive,
and then copy/paste your answers below)
Social Network Analysis and Visualization Checklist
(Steps)
Step 1
Select two social networks
(with >5 people in each) to compare (Q1)
Describe these people and why you think their
networks are worth comparing
Step 2
Generate 5 questions, and
pick one you would like to ask a person in each network (Q2)
Step 3
Use an online or paper-and-pencil
survey to collect the matrix data (Q3)
Step 4
Prepare your 2 matrices for
importing into the NetVis Module (Q4)
Step 5
Login with username and password,
and create your 2 complete networks (Q5)
Step 6
Display your complete networks
to verify that they were imported correctly (Q6)
Step 7
Analyze your complete networks to compute centrality, density, and
transitivity (Q7)
Compare analyses of the two networks for your question
Step 8
Visualize your social network data to see representations of the relations
among people (Q8)
Compare visualizations of the two networks for your question
Step 9
Tell a story that explains the differences you observe (Q9)
Manipulate your visualizations to highlight these differences
Step 1. Select two social networks
As with laboratory experiments or program evaluations, it
is important to make comparisons when analyzing and visualizing social networks.
However, the comparisons must be meaningful. For example, if you wanted to study
the social networks of students looking for summer jobs to
determine what types of networks resulted in better job offers, it is necessary
to survey students who found bad jobs as well as those who found good jobs.
Similarly, if you wanted to study the social networks of senior managers who desired
informational support to determine what types of networks resulted in better support,
it is necessary to survey senior managers who found little support as well as those
who found a lot of support.
Use your personal experience to select two
social networks that represent (approximate) ends
of a continuum, such as high and low performing work groups, successful and
unsuccessful sports teams, effective and ineffective organizations, inactive
and active research labs, productive and unproductive employees, and so on. The
one restriction is that there must be at least 5 members in each social network
(and not more than 10-15 since a person will need
to answer questions about all other people in the network). It is up to you
to decide which two social networks you would like to compare, though keep in
mind that you should try to collect data from a member of the network
(so selecting two networks you are familiar with and have access to may not be such a bad
idea!). It is also recommended that you identify two social networks of the same size,
but sometimes this is not possible for the comparison you want to make.
(Q1) Your first step is to describe the two social networks you selected and
explain why they make for an interesting comparison. Be sure to include the specific
variable you plan to compare (e.g., success, effectiveness, productivity)
as well as the size of the social networks (e.g., 5-10 members).
Step 2. Decide which question
to ask people
Now that you have two social networks to
compare, you next need to think about which factors could possibly distinguish
between them. Is it communication, information, trust, influence, help, intimacy,
self-disclosure, support, resources, identification, or some other factor? Brainstorm
5 questions that reflect your hunch for what would differentiate one social
network from the other. These should be closed-ended questions with ordered
response categories, such as 1-strongly favor, 2-somewhat favor, 3-neither favor
nor oppose, 4-somewhat oppose, 5-strongly oppose or 1-never, 2-monthly, 3-weekly,
4-daily, 5-hourly (5-pt scales would work best). Try to make each question as
simple and easy to answer as possible (e.g., How often do you communicate with...,
How much information do you share with..., How much do you trust...), since at least
one person in each network will need to provide an answer regarding all others
in their network.
(Q2) Generate 5 questions, and put an * next to the question you decided to ask
people in your two social networks.
Step 3. Collect social network
data
With two social networks and a good question to ask, it is
time to collect some matrix data. There are several ways to go about doing this:
(1) provide an online or paper-and-pencil survey that has a matrix ready to
fill in, such as "How well do the following people know each other?"
Person A
Person B
Person C
Person D
Person E
Person A
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Person B
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Person C
Very close
Close
Less close
Distant
N/A
Very close
Close
Less close
Distant
N/A
Person D
Very close
Close
Less close
Distant
N/A
Person E
(2) provide an online or paper-and-pencil
survey that has the question at top with a list of the people in the social
network below along with a line next to their name for writing in the response
to the question. For example:
(1-very close, 2-close, 3-less than close,
4-distant, 5-N/A)
How well does Person A know:
Person B ______
Person C ______
Person D ______
Person E ______
How well does Person B know:
Person C ______
Person D ______
Person E ______
How well does Person C know:
Person D ______
Person E ______
How well does Person D know:
Person E ______
(3) provide an online or paper-and-pencil
survey that combines features of 1 and 2
(1-very close, 2-close, 3-less than close,
4-distant, 5-N/A)
"How well do the following people know each
other?"
Person A - B ___ Person A - C ___ Person A - D ___ Person A - E ___
Person B - C ___ Person B - D ___ Person B - E ___
Person C - D ___ Person C - E ___
Person D - E ___
The important thing to remember is that
people are answering a question about all relations in their social network.
This way, you have an estimate of all possible connections, not just the
connection between the survey respondent and others.
For this activity, you only need to collect data from one person in each network, though if you
do decide to collect data from more than one person, it is a good idea to average
the responses in the matrices "(Person A + Person B) / 2" and round up to the nearest whole number.
(Q3) Using an online or paper-and-pencil
survey, collect matrix data from people in your two social networks and describe
the procedure you used (e.g., which survey method you used, how many surveys
you sent out, how many surveys were actually completed, etc.).
Step 4. Prepare matrix data
for use in the NetVis Module
Your matrix data (whether through the web, email, or paper),
should look something like this for the two networks:
(a)
(b)
4,5,1,2
1,3,5,1
1,2,3
4,2,1
1,4
1,3
4
2
Now it needs to be processed in a format that is recognized
by the software. There are three parts to making this work:
(1)Each matrix should include alphabetized, comma-delimited variable
labels in the first row (give each person in your dataset a name)
(2)Each matrix should include comma-delimited values
in remaining rows (place the data from each person on a single row below the labels)
(3)Each matrix should be symmetrized, including zeros on the
diagonal (people don't usually talk to themselves!)
After following the three parts from above, stacks the two datasets
on top of one another like this:
Person A, Person B, Person C, Person D, Person E
0,4,5,1,2
4,0,1,2,3
5,1,0,1,4
1,2,1,0,4
2,3,4,4,0
Person A, Person B, Person C, Person D, Person E
0,1,3,5,1
1,0,4,2,1
3,4,0,1,3
5,2,1,0,2
1,1,3,2,0
(notice that the values
from the lower-left triangle reflect the values from the upper-right triangle -- a symmetrized matrix)
(Q4) Copy and paste the matrix
data below for your two social networks
Step 5. Create your social networks
in the NetVis Module
If you are not logged into the NetVis Module, use your username
and password to login again. You should be at the socialnetworks.php
page, which serves as a hub for displaying, analyzing, and visualizing data.
Since you have not created a network yet, there will not be any social networks
showing. Please click on "help" if you want to learn more.
After clicking on "Add Data", you should arrive
at the data.php page, where you can 'Create a File' or 'Upload a File.'
Click on 'Create a File' and enter the name of each social network in your data separated by a comma, such
as 'goodnet,badnet' (again, click "example" to double-check that your names are
in the correct format) followed by a meta tag (such as 'Communication').
Then, copy/paste your two matrices into the textbox. Since your data above are already in csv format
(click "example" to double-check that your matrices are in the correct format),
you can submit your information. All other settings can be left as default.
(Q5) Describe the settings you
used to import your two social networks, noting any changes you may have made
to the default settings
Step 6. Display your social
networks
After creating a data file, you should be
back on the socialnetworks.php page with many options. The first thing
you want to do is display your complete networks to verify that the data were
imported correctly. Click on "Display All". Notice that the matrix data are
available as csv files on the display.php
page.
(Q6) Copy and paste your csv
output (matrix data) for each social network
Step 7. Analyze your social
networks
From the display.php page, you can
click "Analyze All" at the bottom to directly compute Centrality (how clustered
in the center), Density (how dense overall), and Transitivity (if A->B, and
B->C, does A->C). Notice that there are three measures of centrality,
which reflects the concepts of Degree (ties into and out from), Betweenness
(bridge among), and Closeness (shortest path). Higher values represent greater
centrality, density, and transitivity, and are commonly used to describe the
structure of social networks. Notice that the values are available as csv files
on the analyze.php page.
(Q7) Copy and paste your csv
output (network analyses) for each social network, and briefly compare the
analyses of the two social networks
Step 8. Visualize your social
networks
You should 'Return to Social Networks,' and
click "Visualize" to directly render an image of the relations
among people in your social networks. The NetVis Module relies on Simulated
Annealing to compute the 'distances' on the graph. For example, if you measure
communication, people closer together on the graph communicate more with one
another and people further away on the graph communicate less. This step should be repeated for each social network.
To save the image, use the 'Print Screen' function on your computer.
(Q8) Briefly compare the visual representations of the two social networks
Step 9. Tell a story about
your social networks
With the analyses and visualization complete,
the critical step lies before you: interpretation. Social networks are valuable,
in part, because they often help communicate a story. Tell an interesting story
about why the two social networks differed, and use your skills at manipulating
the image to highlight these differences. When you have two images that you
are satisfied with (one from each social network), you are ready to begin your story.
In your story, be as descriptive as possible,
including where the nodes and relations are positioned in the image, what it means
when nodes are connected (besides that they had a value higher than the split
value!) or distant from one another. Use the network analyses to support your
story where appropriate, and remember to frame the story in terms of the "comparison"
you set out to establish in the beginning.
(Q9) Tell your story!
You've completed the tutorial! You can now use the NetVis Module for analyzing and visualizing additional social networks!
For further reference, you may want to print out a copy of this page with your saved answers.
