Last chapter, interaction.
Let's just start by taking a look at
a very classic data of this example.
And this is Marey's Paris to Lyon train
schedule, and this comes from 1885.
It's one of those sort of
masterpieces of visualization.
And what I'd like to do is show you
how something even as marvelous as
this can be improved with
something called interaction.
So, in this particular diagram,
the x axis represents time and the y axis
represents the nominal location for
top Paris, bottom Lyon.
And what you can see is,
each train is represented by a different,
a line of a different slope.
So lines that are sloping to the right
are trains that go from Paris to Lyon,
and trains that are sloping up into the
right are trains going from Lyon to Paris.
So something as wonderful as this can be
improved with some very
simple interaction.
So what we've done here is this
is a code written by Mike Bostock
in a very excellent online
graphing library called D3,
and he's added some interactive
widgets across the top.
And this allows us to do things like, well
let's just turn off the southbound trains.
And I'm only interested in not
the everyday but just Saturdays.
And I'm only interested
in the bullet trains.
Oh look,
there's actually none on Saturday.
So I wonder,
I'll have to look at weekdays.
Looks like there's plenty during weekdays.
So you're ability to actually change
what data are being represented
from the entirety of the data
set gives us an incredible power
to see patterns that are ones
that aren't necessarily
visible in the encoding of
the data set in its entirety.
Let's look at some other examples,
and see how our ability to interact
with data also give us an incredible
power to learn things about patterns.
So here is something
called the zip decoder.
It's made by Ben Fry.
It's a wonderfully simple data
visualization where what Ben has done is
he's mapped every, unique postal code
to a dot, in latitude and longitude.
And what you can do is, by clicking on the
map and you enter, everybody just enters
their own zip code, and they're able to
see the, how the zip codes form a shape.
And so for example,
I come from Long Island,
New York, so I would type one, and you
can see here that this is the Northeast.
One, that's Long Island, and
then seven, four, seven,
and you can see we've slowly,
gradually reduced the amount of noise.
But, of course you can
enter whatever you want and
you can see here quickly
that if you just type nine,
that this is the West Coast, and
that eight is slightly more inward.
Seven is Texas in the South.
Six, five, four, so
you can see there are different patterns,
that in the entirety aren't visible.
But by our ability to interact,
what we've really done here is
something called dynamic filtering.
And what we're doing is we're maintaining
the representation, but we are removing or
highlighting data depending on
commands that we give the computer.
And so here when I'm saying to
show me just these, what it's
really doing is making more salient the
data that I'm interested in looking at.
Another example of this type of filtering;
here is the different types of travel
you know, airline travel
between where is it exactly?
[LAUGH]
Between somewhere and somewhere else.
But the main point is that what you can
see here is the top is a one chart,
a histogram, showing arrivals by
time of day and this range of date.
And what we can do here is if we increase
the date, what you can see is we ad and
you'll see the time of day that
histogram is getting the shape,
the shape of the histogram is
changing based on the actual data.
And we're changing the entirety of
the data that's being included, and
then what you can do is using the cursor,
select just a sub sample of these below.
So now what I'm doing is looking at
data between January and February, and
just at, you know, between six and
10 in the morning.
And I can now look at
different flights here.
So if I'm interested in seeing, oh well
what about flights that are, you know,
in February and that are in the afternoon.
We can see here that this is
the list of flights below and
up top,
we can see the up time the arrival delay.
So for example if I want to pick flight,
of these flights,
the ones that are the most on time,
I can sub select again.
So this ability to filter, right,
the underlying representation named is
the same, and
the visual presentation is the same.
All we're adding is this layer that's
saying, show me some of the data,
but not all of the data.
And, because you're showing the data, that
maps to the questions that you're asking,
you're actually able to understand
a great deal about the data.
Another wonderful example of this
kind of dynamic filtering is a data
visualization called five years of info
aesthetic which takes the info aesthetic
visualization blog and every single
one of the features from that blog.
And so for example, I can and you can see,
it organizes them by category.
And if I click on one,
it will show me what categories that
particular visualization belong to, and it
will actually get an image that represents
it and show me the ability to go there.
And now if I was interested in sub
sampling I could actually say well how
about the, this particular visualization
well this one is an aesthetic or
info graphic.
Oh, well I see how this works.
What if I'm only interested
in seeing the ones with art.
And so what happens is, the art becomes
the top level representation, and
all of the, the headlines that don't
include art become diminished.
So, now I'm actually able to sample,
through time, the different visualization,
the different features that include
art as their top level component.
So what you can see here is
this is a fascinating way to,
by using this idea of dynamic filtering
and changing what's being presented,
you're able to give the individual
this capability to see into the data.
So visualization excuse me,
interaction can be a hard thing to fathom.
So I look to Ben Shneiderman's
example of what is it that we
should do in order to make interaction
with visualization as simple as possible.
And he has three wonderful guidelines.
Overview first, zoom and
filter, details-on-demand.
So, let's go back to look at this.
When we start with the overview,
what we're able to do is see the entirety.
Wow, there are probably 500 of these,
you can see the number and
the distribution of,
over subjects on the right.
Overview is what you get
when you first come.
Zoom and filter.
So, I'm able, I'm able to filter and
I'm able to zoom by seeing, well actually
I'm not able to zoom in this one, but.
And then details on demand, so
when I'm interested in finding out more,
I can enter very specifically,
I want to see this one, and
then it gives you some of
the much more detailed layer.
So, you see,
by layering the visualization and
allowing people to stand back and
see the big picture or to zoom in, and
see the individual data points
you're taking advantage of one of
the things that interaction can do that
other types of visualization simply can't.
Looking so last
something that I think is just a joy I
recommend everybody doing is kind of fun.
It's something that was come
up by Martin Wattenberg and
it's basically called
the baby name generator.
Or it really is, is a statistical
evaluation of names mentioned
in the United States Census
from the 1880s to today.
And it shows the top 2,000
most popular names and
what you can do is track the historical
popularity of different names.
So I'll write here for example if I
want to look at Scott, I would type in S,
it'll filter down on different
names that start from S.
And what you can see here is, Scott wasn't
actually very popular in the United States
until the 1930s,
achieved peak popularity in the 50s and
60s, and then is now actually
not a very common name at all.
And then if you look at something like L,
you can see names like Larry,
more classic American names or
a name like Linda which had a burst
in popularity between the 30s and
the 60s and is now represents a much
smaller percentage of the population.
Fine interactive date visualizations.
I think which we'll find
is that interaction is
one of the hardest parts of
data visualization to master.
But it's also one of the richest,
because it's the part that allows
you to dive into a visualization and
extract answers, that
are potentially harder to gather
when looking at your data
from a single perspective is
challenging because of its density or
perhaps you're not even sure what
you're looking for right yet.
So, I recommend, tremendous,
well actually one other thing.
There's I'll just put
this teaser at the end.
Last year myself and
a number of colleagues gave a number of
keynote lectures on the subject
of What is data visualization.
I think you'll find that there's
a really wonderful cross section
of thoughts about what visualization is.
If you look at the handouts
from the lectures,
each of the links will be in there,
so you'll be able to click them and
explore each of the talks
is about an hour.
So enjoy them.
I hope this was interesting,
and, you know, think visually.
Thanks very much