Histograms in R

NARRATOR: Hi! In this video,

we are going to be talking about
graphical data summary

for numerical data.

We are going to be, um,

using the iris dataset,

for this.

And the first graphical data summary
we're going to talk about are histograms.

So, if you remember from the iris dataset,

there were four different
numerical variables you could choose from.

Sepal length, sepal width,
petal length and petal width.

And I'm going to use
the petal length, um, variable today.

So, to create a histogram of a--
of a numeric var-variable,

you will use the function, "hist()."

And then, you can just go ahead and put in

whatever data you want
to produce a histogram for.

Since this is already
a numerical variable,

it's good to go!

All you need is just this.

And here you can see a histogram,

of all the various petal lengths.

There's a lot of them that are
in between one and one and a half,

and, yknow, fair amount over here.

Definitely not really
normally distributed,

not like a bell curve.

Kind of, got a couple of peaks.

So, if you want to customize
your...histogram,

here are some additional arguments
that you can put in, to make it,

more pretty, I guess.
[chuckles]

So first one is,
similar to categorical data

—when we were doing the graphical.

You can create a title.

So there is a default title here,
"Histogram of..."

But if you don't really want
this R syntax to be showing,

you can totally change it.

So...

Let's do...

"Histogram of Iris Petal Lengths."

There we go, much better.

We can also, um, add,

x-labels for the x-axis,

so if we didn't want
to have that there, as well.

You can also change the y-label,

but I'm going to keep it just like that.

So, and you know,
these are all available to you,

but you don't have to change all of them.

Just kind of whatever you feel, so.

I'm going to call this, "Petal Length."

There we go, looks much better down here.

You can change the limits
of your x-axis and your y-axis.

So obviously, with the Y,

it's not quite reaching
all the way up there.

So I'm going to change that.

And then, I like to always have
a little buffer on the sides, too,

so maybe we could have it
go from zero to eight or something.

So for the x-axis, I'm going to change
my limits to go from 0 to 8.

And my y-axis, I'm going to have
the axis go from 0 to 60.

Now,

you just kind of just play around
and find out which one looks good,

and kind of when you figured out
what you like the best.

There really is no right or wrong, um,

magic number.

Just kind of--to make it look nice,
is all we're really kind of doing.

Alright, so now you can see here.

Maybe 60 was a little too much, but.

Eh, it works.

But now you can see there's a side,
buffers on the side.

You can see it easily,
where it easily starts and ends.

So, I probably actually might take
that in a little more

and squish that down
a little more, but, y'know.

For our purposes, it's probably just fine.

There is another argument that you can do
that's new that's called "breaks,"

where "breaks" kind of specifies
a suggested number of bins.

It can't always perfectly say like,

so if I want to say breaks--

And by bins,
I'm meaning how many, like,

boxes it's using and how many--
to create the histogram.

So if I were to say like, eight,

it's going to try and do
its very best to create it

so that there's only eight boxes.

So there's 1, 2...

—there would technically be 3—
4, 5, 6, 7, 8, 9, 10, 11, 12, right now.

Let's see how well it can do with eight.

If you specify seven,
it might do six or eight,

just kind of depending on how the data is.

it does its very best to match
whatever number you put there.

So you see here,
we got one, two, three, four, five, six.

So yeah, so it--

It definitely made it so that the shape
of the graphic looks different.

The, y'know the--bars are wider.

But it didn't quite do it
exactly as eight.

So that one is, uh.

You just kind of
have to do your very best.

If you want to learn more
about how you can kind of be more specific

on how you can set the actual breakpoints,

you should take
Statistical Visualization.

Or, you can research how to do that.

Or you can use the "ggplot" package,
and that's pretty great as well.

Another thing you can do is,

you can say labels is equal to true.

And what this will do is that
it will put the counts in of--

in each category on top of the bars.

So you can see that there were
50 in between one and two,

1 between two and three,
15 between three and four, etcetera.

Another...
And then the last argument,

—oh actually, the second-to-last argument
is you can always change the color.

And I'm going to pick
a color called sky blue.

Doesn't that look pretty?

Alright!

The one other thing we will talk about

is sometimes, you could have
a frequency histogram

where you just say how many were
in between one and two, two and three

and put them in there.

Or you can do a, um,

based off of kind of,
probability densities.

So meaning, this is, um...

probably, y'know,
what proportion of the data lies in here.

So instead of having these
be frequency numbers,

they would be, like,
proportion of the whole.

So, y'know, this has got quite a bit,

so this could be like, y'know,
almost a third of the data.

Or a quarter of the data
could be all in just this section.

And so you can also make
a histogram be like that as well.

So I'm going to...

copy this.

And make it be very similar,

but I'm going to kind of just
say it's a density function now.

I'm going to temporarily take out
the "ylim" argument.

And then we'll just kind of see...

how that works right now.

And then the only other argument
you would need to change

is just say "freq," or frequency, F-R-E-Q.

Say it's equal to false, meaning,
do not make it a frequency table.

And see what that does.

So now you can see
that this is about a third of the data.

Oh, it's getting cut off a little bit

'cause we'll wanna change our Y limits.

Notice it's not going from 0 to 60,

and because these are all now
percentages or proportions.

So, I'm going to make my Y limit be...

From zero to one which is the max
that it kind of can be, so.

Yeah that was probably a little too high,
but I mean it still gets to point across.

So as you can see, almost a third
of the data is all contained here.

Then we've got, 0.7% of the data is here.

If you were to add up
all of these numbers together,

it would sum to one.

Just like the, um.

A probability distribution function.

The area under all of its curve,

in its--in it's, kind of,
in its range is, um,

needs to sum up to one.

So this is a way that you can kind of
show that, with your own data.

Now, if you want to create a histogram
of the petal length variable,

but remember that there were
three different species as well.

There was setosa,
versicolor, and virginica.

If you wanted to show a histogram
of just the petal length for like,

just the setosa variable,
[stammers] setosa species, excuse me.

There isn't a quick and easy way
to do that.

You just kind of have to create
a separate histogram for each, so.

Say like, you wanted a histogram
of just the petal length for setosas,

just for virginicas,
and just for versicolors.

This is how we can do that.

I'm going to copy and paste
some code that I have,

just so you can see.

So, histograms
for the petal length variable,

I didn't quite do all of the different
specifications and customizations,

but I did enough to make it look good.

So this is what it looks like
for all of the petal lengths,

not for separated by species.

If you want to separate them out
by species, you can do that

by using bracket notation
and logical operators.

So bracket notation
is how you subset data.

So, what we're doing here
is we're saying,

"I want the petal length—the variable,
the numerical variable—to be plotted,

but I only want the values
that meet a certain criteria.

And the criteria is, we want it--

the only the ones
where the species is equal to setosa.

So this will now plot a histogram
of all of the petal lengths,

but only for the ones that are setosas.

So what's interesting here
is you can see that

all of the setosas have
really small petal lengths.

And I do the exact same thing, um, here.

I just change each of these
to versicolor and virginica.

And the colors, too.

And I have it so that they're all
on the same scale as well,

so that it's easier to compare.

That's also really important
to make sure you do.

If you're going to make
comparative histograms,

make sure they're all on the same scale.

Here's what versicolor would look like.

So the versicolor ones are all right here.

And the virginicas...are--
are the biggest of the petal lengths.

So as you can kind of see here,
they start at about 4.5.

These start--they have some in the 4.5,
but it's definitely in this middle range.

And then the petal--
the setosas are all the way over here.

And so if you compare that to this,
this is obviously the setosas.

These are the versicolors,
and these are kind of more the virginicas.

There is a way to also, um,

in uh, to possi--

in other packages and things in R
for you to even specify it even further.

So you could maybe even show that,
this is setosa, versicolor, virginica.

But for right now,

we're just gonna show you
how you can do each one separately.

This is probably the easiest way
to go about doing that.

Alright, that is everything, um,

you should need to know
about creating histograms

for your numerical data in R.

And we will see you in the next video.