R Classes and Methods (Part 2)

Loading...

En provenance du cours de Johns Hopkins University

Developing Data Products

1328 notes

Johns Hopkins University

Developing Data Products

1328 notes

Cours 9 sur 10 dans Specialization Data Science

A data product is the production output from a statistical analysis. Data products automate complex analysis tasks or use technology to expand the utility of a data informed model, algorithm or inference. This course covers the basics of creating data products using Shiny, R packages, and interactive graphics. The course will focus on the statistical fundamentals of creating a data product that can be used to tell a story about data to a mass audience.

À partir de la leçon

R Packages

In this module, we'll dive into the world of creating R packages and practice developing an R Markdown presentation that includes a data visualization built using Plotly.

R Packages (Part 1)7:11

R Packages (Part 2)14:59

Building R Packages Demo18:00

R Classes and Methods (Part 1)13:49

R Classes and Methods (Part 2)11:19

Rencontrer les enseignants

Brian Caffo, PhD
Professor, Biostatistics
Bloomberg School of Public Health
Jeff Leek, PhD
Associate Professor, Biostatistics
Bloomberg School of Public Health
Roger D. Peng, PhD
Associate Professor, Biostatistics
Bloomberg School of Public Health

So here's a, simple example of a S3 class and method.

So in the, in the top I'm generating some random, normal data.

And, I'm calculating the mean.

So here the mean is -0.0307.

So that's not particularly complicated.

But here's a lot of stuff going on behind the scenes.

And so, first of all the class of x, which is this random normal vector, is numeric.

So I called the mean, which is a generic function on this numeric vector,

it will search to see if there's a specific mean method for numeric objects.

But there is no as it turns out there is no

method for a numeric object, so we'll call the default function.

For mean, which calculates you know, the sum of

all the elements, and divides by the length the vector.

So you can look at the default method for the mean function by using

the get SE method function, and you can see that there are other arguments.

There's this trim argument, and a .rm, and as you go through,

the function, you can see that it checks a number of things.

It look, if you're doing a trimmed mean, it takes care of that first.

And then, and then ultimately at the very bottom there, it calls

some internal C code that actually

calculates the meme for efficiency purposes.

[BLANK_AUDIO]

So that's sort of.

The meme is a simple example here.

Is, is.

And as, a slightly more complicated example I've

generated some, random data in a data frame now.

And I'm S applying The mean function over the data frame.

So one of the important things to known about data-frames

is, that each column could potentially be of a different class.

So you could have the first column be numeric like we have here

and then the second column be integer so when we add supply over the

columns and we call the mean function for each column the mean will, will

protect the class of that column and see if there is an appropriate method.

So for the first column it's numeric It will check for

a numeric method, there isn't one, so it will call the default.

And then the second column is actually integer, so it will check for an integer

method, again there's no integer method, so it

will call the default for that column too.

So, for, if you have a large data frame

and every column, is a, is a slightly, is a

different class, the mean function will check each column

to see if there's an appropriate method for that class.

So in some cases it's possible to call the, a method directly, so

you may suffer some S3 methods are visible to the user, so you make.

So, for example the mean.default function, is you can

possible to call that directly without calling the generic function.

But in general the rule is you should never call methods directly.

You should always call the generic function and

let the, the appropriate message be dispatched automatically.

And this, this results in cleaner code, And and it's slightly more robust.

That way the name of the, of the method switches so if it

changes a little bit you won't have to worry about the underlining details.

With the S4 system this isn't a problem because you

cannot call the methods directly for the most part at all.

So one last method, example for an S3 Class Method so here is the plot function.

And again I'm generating some random normal data so when I

plot it again it looks for a numeric method for plot.

And since there isn't one it just calls the default method for

plot, and it makes a little scatter plot, as you probably would've expect.

However, this is a slight variation.

I'm generating some random normal data, and then I'm converting, it into a time

series object, so I use the as.ts function to convert it into a time series.

Now, you'll notice I call plot, I call plot the exactly

the same way as I called it in the previous slide.

But now you can see the plot is totally different.

Instead of a plot with a bunch of circles on it it actually creates

a kind of type series kind of plot where all the, the lines are connected.

And also you'll notice on the x axis, there's a label on it for, called time,

which is different from the x axis that

was called for the default function in plot.

So here you can see that there is a

special plotting method for ts objects Or time series objects.

And that plotting method is actually being called here and not the default method.

So, if you want to write new methods for new classes, if you often create a new

class, if you represent a new data type, you'll

probably end up writing methods for printing or showing.

You'll probably, may write a method for summary.

You'll often, you'll write a method for plotting because there

won't be a plot method for a new type of data.

And so there are a couple, there are basic, two basic

ways that you can extend R with the classes and method system.

You can write a new, you can create a new

class and then write a method for an existing generic function

like Print or Plot Or you can write new generic

functions, and new methods for those generics for your new class.

So, for the rest of this lecture, I'm going to talk about S4 classes.

And, and so, one of these questions, is, you

know, why would you want to create a new class?

Right.

So what's wrong with the vector in lists,

and numerics and integers and logicals, et cetera.

While it's possible to survive on just the basic data types often it's easier

and more compact to think at a higher level about, more complex data types so,

for example, you might want to represent new types of data like gene expression

data Spacial temporal data, hierarchical datas that

you might want to represent a sparse matrix.

These are all data types that don't exist in r.

And so, you have to create a new class to represent them.

There are often new concepts that haven't been thought of yet.

So, if you can think of, you know, there's a linear model type of

class, but you think of point process model is a different kind of class.

Mixed effects model is another type.

Things like that that haven't been thought of

that haven't been kind of implemented yet in R.

And also you might want to hide certain implementation details from teh user.

And so you, you may be able to represent a new type of data using lists.

And vectors.

But then there maybe a lot of ugly details that are exposed

to the user that you would rather not, the user know about.

And so, And.

And one thing to, to emphasize, that when I say creating a new data,

type I don't mean, necessarily mean that the data type has never been seen before.

All I mean is that it's not known to R

and R doesn't have any special handling for those data types.

Types.

So, creating a new class is done with the set class function.

At the very minimum you can get away with

just specifying the name of the new class, but often

when you specify a new class there will be

data elements associated with this class, those are called slots.

You can define methods For this class using the set method function.

So, at a loose level, you can think of a class as like a list.

Every class has a bunch of slots.

So, these are kind of like elements of a list.

But each slot, it's a little bit more specific,

because each slot is an object of a certain class.

So you can't just put arbitrary data into any slot of a class.

You have to put the specific type of data, into each slot of a class.

So, for example, here I have a simple example of creating

a polygon class, so there's no polygon, data object in r.

So there, you can think of a variety

of ways in which you might represent polygon data.

So, you can think of a polygon, as a set of vertices, and a set of

or it's just a set of vertices and

in thinking you can have lines connecting the vertices.

So here I'm creating a polygon class.

And in the representation, I I create two slots.

One is called x, which contains the x coordinates of all of the vertices.

And then another slot is called y, and

that contains the y coordinates for all the vertices.

You can see that, so I've specified that the x and the y slots are to be numeric.

So if I create a polygon object the x

and the y slots have, have to have numeric data.

They can't have characters, they can't have integers, etcetera.

So now I've created a polygon class the next thing I probably want to

do is create a method for, for example, for plotting this, a polygon.

And so when you call the set method function you have to specify generic

functions such as plot, and then you have to specify something called a signature.

And so the signature is basically the, the set of

classes on which the generic on which the method will operate.

So in this, case I'm going to want to

create a method for the plot generic function.

The signature is basically going to be this Polygon Class that I just created.

So here's my method for the plot generic called the set method function.

The first argument is the name of the generic, and the second argument is the

signature, in this case, it's polygon and so you can see in this function it

takes an X and a Y argument and then the dot, dot, dot is for

kind of other arguments And so you can see that within so the x argument.

Is just the polygon object.

The y argument is missing there's no y argument here.

And so I can see that within the function I actually called plot again.

Right so here I'm creating a method for the polygon object.

And within the method I actually call plot again.

But that's going to be the default method for plot.

Because I'm just plotting some numeric vectors in that case.

So what this function does is it plots the vertices.

And then it sits, to set up the plotting window.

And it kind of wraps the vertices around themselves so that they kind of connect.

And then it uses the lines functions to to connect, connect, the dots.

And then after that, it makes my little polygon here.

And so that's the plotting method for polygon objects.

I know, notice that I, the plot, since, since plot already existed

as a generic function, I didn't have to create a new generic function.

And notice that when I want to access the slots

of an object in an s4 class, I use the at

symbol to so it's kind of like a list but

instead of using the dollar sign I use the at symbol.

So after I call the set method function, if I call

show methods on plot you can see that the the polygon method

has been added to the list here and then in addition to

the polygon method there is any method which is the default method.

So here I am creating.

I'm using the new function to create a polygon, an object from the polygon class.

I create the x vertices and I create the y vertices and now

when I call plot on it you can see that it doesn't just

call the default plot method, it actually calls my method that I just

created and you can see it draws this little triangle by connecting the dots.

So, that's a very simple example on how to

develop, create a new class and develop some new methods.

It's a in general developing classes and methods is

a very powerful way to extend the functionality of R.

And so just, so just to summarize, their classes define new data

type so they, they, they allow R to represent new types of data.

Methods extend generic functions to, to specify the behavior

of generic functions on these new classes that he developed.

And as you develop these new data types and these

new concepts and make a familiar to R, the classes and

methods give you a way to develop kind of an easier

interface for users to kind of interact with new types of data.

And so it's really a handy way to

kind of create new ,to allow users to interact with

new kinds of data without, without having to

get bogged down, a lot of the implementation details.

And one of the most popular ways to kind of make

new classes and methods available to users is through R packages.

And so, most commonly you'll see these

kinds of things embedded within an R package.

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2018 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play