Now before we start using Gephi, let me
tell you about the software tools that we
will and won't be using.
In this class, we'll be using Gephi for
visualization and for calculating simple
network metrics.
We'll be using NetLogo to understand the
dynamics that shape network formation and
also to understand processes occurring on
networks.
And finally for those of you doing
programming assignments we'll be using
iGraph for more sophisticated, more
flexible calculations.
There are actually a lot of alternatives,
and let me tell you about some of them,
because once you know what capabilities
you like, you can readily find them in
other software that may be more
well-suited to your projects.
So there's the software package called
Pajek which means spider in Slovenian.
It has been developed for a number of
years, it's a very mature software
package.
It has very extensive functionality which
you access through drop down menus.
And the reason why we're not using it even
though I have used it when teaching this
type of class in the past is because it
only works on the Windows platform and
some students have had trouble using
Windows emulators on other systems.
So I just made the choice to switch over
to platform-independent software.
However Pajek might be a very good choice.
In addition, there is a book, a textbook,
called Social Network and, Exploratory
Social Network Analysis with Pajek, which
guides you through many analyses that you
can do with that software.
We also will not be using UCINet, which
again is fairly mature, or very mature.
It has extensive functionality, mostly
focused around sociology so, I think until
recently UCI Net did not really scale up
to large networks.
Unfortunately it's, it's also Windows
only, and the license may cost you a
couple hundred dollars if you're not a, a,
a full time student, or somehow affiliated
with the university.
We also will be using NodeXL, NodeXL is a
relative newcomer.
Its social network analysis neatly
integrated into Excel.
It's still, I think, very much in beta,
it's being developed.
However, it's free.
And if you're, are already an Excel user,
then this may be a very friendly way to do
social network analysis.
Again, it's, it's actually Windows only,
only, the, Windows version of Excel
supports NodeXL, and that's why we're not
using it.
If you're a programmer, there's an
excellent alternative called NetworkX.
It has extensive functionality and it's
actually built upon, Fortran and C
libraries, which are optimized to scale
for very, very large matrices.
And so you can do calculations for very
large networks.
It's, on top of that it's open source, so
that's definitely something to consider if
you have this background.
There are many, many other alternatives,
most of them are in various phases of
development one very mature package is the
social network analysis package for R, it,
it has a kind of statistics heavy duty
functionality for example, p-star models
which we won't even get into in, in this
course, are supported in this social
network analysis package.
In addition, if you are interested some,
p-star models are used to model networks
as they evolve to time.
Just a hypothesis about, what's driving
edge formation.
And if you're interested in visualizing
networks over time with kind of an
integrated P star model functionality,
there's the package SoNIA, that, that you
can use for this purpose.
Now let's get to our main task for this
lecture, which is to work with Gephi.
So what I'd like you to do is to download
Gephi.
It's free, it should work on any platform,
and install it.
Once you've installed it, download the
data file dining.gephi from the Coursera
folder for this lecture, and we can start
to play.
Now that you downloaded Gephi and
installed it, we can start to play with
this dining.gephi data file.
Now I've opened it recently so it's shown
in this helpful welcome screen but perhaps
you don't have that so you can go to File,
Open, and navigate to the directory where
you've saved it and then just double click
to open it.
And here we're waiting for it to load and
pretty soon it's going to tell us, for
example, the number of nodes and edges.
It's going to tell us there are 26 nodes
and 52 edges.
What kind of data is this?
Well, this was data gathered in a girls'
dormitory back into the 1940s, where the
girls were asked to name their first and
second choices for who they would like to
sit with at dinner.
So, this is the dining table partners'
data set.
You, you can see a random layout here, and
I've done this on purpose to demonstrate
how even with such a small network as this
one, a random layout can really obscure
whatever insights you're trying to see.
Now, Gephi helps a little bit.
Here we can tell for example, that this
girl is relatively popular because lots of
other girls have named her as their first
or second choice.
Now, if we want to know which girl exactly
this is, what we do is we select this
little arrow.
Plus question mark.
It's actually an, brings up an edit window
over here, and now when we left-click on
this node, we find out that this girl's
name is Eva, in addition to some other
properties.
We also may want to know the overall
structure so we can select a layout.
And here I went to the layout panel, and
I'm going to choose a layout.
Force Atlas two is one of my favorites,
but Yen Phen Hu is also very fine.
It turns out it doesn't matter too much
when the network is this small.
It's more with large networks that you run
into utilization difficulties.
So, let's try Force Atlas two.
And I'm going to click on run.
There're all of these different parameters
which you can tweak and, and play with
that have to do with how the model's
running.
Essentially, all of these different layout
algorithms want to place nodes that are
linked close together and ones that are
not.
It wants them to repel from each other.
So we're going to say run.
And it ran, it's doing a little bit of
spinning.
Let's see if it'll actually spin into a
good position.
Oh, indeed, very nice.
Okay.
Let's stop it right there.
And you may notice that now some nodes are
out of use, so you can click this little
magnifying glass that will center it on
the graph but it may end up looking a
little bit too small.
So I'm going to use the track pad or you
can use your mouse wheel to just zoom in.
And then I'm going to right click and drag
and zoom in some more.
And this I think shows off the layout of
the girls' preferences a little bit more
clearly.
You may want to tweak things just a little
bit.
I'm going to select the hand to drag the
node just here.
It wasn't quite clear who was, who was
picking whom, so maybe I just move it a
little bit, right, so a, a little bit of
minor adjustment is no big deal.
If you're moving things around because you
want them to look a certain way to prove
some point, then I wouldn't, I wouldn't do
this.
But just for us and wanting to see who
exactly has picked whom, I think it's
perfectly fine.
Okay.
Some other things that you can do is you
can change the color for all the nodes.
So, if we want them all to be red, and
then we, and so I right-click to choose
the color, and then I left-click to set
them all to red.
I may want to, yes okay, these were a
little overlapped.
So we just want to avoid overlaps where we
don't know where there's an edge.
Is incident on the node, or the node is
over the edge and it's actually obscuring
it.
We can also adjust the size if they seem a
little bit big.
I'm going to right click on the size.
Well actually, 1.0, that's already quite,
quite fine.
I guess you could try a fraction of the
size.
I'm going to left-click and now they're
tiny, so let's go back to 1.0.
Inset that.
Oops, alright.
Let's see.
Five?
Apparently it wasn't five before.
Okay.
There we go.
So they're a size five.
We may find this more informative if
instead of having to click on this edit
attribute where we then click on
individual nodes and find out, hey, this
girl is Hazel.
And this one is Maxine, this is where I'm
seeing all this info after I've selected
the Edit tool.
Perhaps if we just show the labels for all
the nodes, so I'm just gonna click on this
T down here that says show node labels.
And they are way too big so we're just
going to have to try and scale then down a
little bit.
There we go.
And, we can go back to the layout and say
label adjust.
Try to move these around so that the
labels don't overlap as much, and it's
actually done it.
I don't know if you could see it, it
popped up very quickly.
And now, all the labels are fairly
visible.
Another thing that we may want to do is to
color the edges differently depending on
whether this is a girl's first or second
choice.
So I'm going to go to the partition tab,
click on it.
I'm going to click on edges and I'm going
to click this little reload.
Load the different partitions or actually
characteristics of the edges that are
discreet categories and I'm going to click
on label.
And this has whether it was first choice
or second choice, so I am going to click
apply.
And now we see all of the second choices
in blue and the first choices in red.
And I will start here because we need to
explain some additional concepts before we
can use Gephi, oh, but actually, let me
show you.
So, we're we are right now in, in overview
mode where we have all these different
tools that we can use.
We can also click on data laboratory, And
in data laboratory, we have a spreadsheet
of all of the node information, and if we
click on edges, we have all of the edge
information as well.
And finally in Preview we can preview the
visualization that then we can export.
So I'm going to click Refresh now to see
it.
I can, I should probably see the node
labels so I'm going to click that and hit
refresh.
And then, I'm going to unclick curve so
that we get the arrows for the
directionality of the edge and perhaps I
can also well change the background, in
this case maybe I want a blue background
or something like that.
No, not a blue background.
Let's go back to white.
And then if I wanted to export this I
would click this button down here and I
could choose whether to export it as a pdf
or a png or an sv2 file.
I'm not going to do that right now because
we have some more work to do.
So I'm going to go back to overview and
stop here while we learn a few more
concepts and then we'll be back.