knitr (part 4)

Loading...

From the course by Johns Hopkins University

Reproducible Research

2610 оценки

Johns Hopkins University

Reproducible Research

2610 оценки

Course 5 of 10 in the Specialization Data Science

This course focuses on the concepts and tools behind reporting modern data analyses in a reproducible manner. Reproducible research is the idea that data analyses, and more generally, scientific claims, are published with their data and software code so that others may verify the findings and build upon them. The need for reproducibility is increasing dramatically as data analyses become more complex, involving larger datasets and more sophisticated computations. Reproducibility allows for people to focus on the actual content of a data analysis, rather than on superficial details reported in a written summary. In addition, reproducibility makes an analysis more useful to others because the data and code that actually conducted the analysis are available. This course will focus on literate statistical analysis tools which allow one to publish data analyses in a single document that allows others to easily execute the same analysis to obtain the same results.

From the lesson

Week 2: Markdown & knitr

This week we cover some of the core tools for developing reproducible documents. We cover the literate programming tool knitr and show how to integrate it with Markdown to publish reproducible web documents. We also introduce the first peer assessment which will require you to write up a reproducible data analysis using knitr.

Coding Standards in R8:58

R Markdown Demonstration7:23

knitr (part 1)7:04

knitr (part 2) 4:10

knitr (part 3) 4:46

knitr (part 4) 9:29

Meet the Instructors

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

So the processing of the knitr document is here, is, happens in a certain way.

First the RMarkdown document is processed and creates

a markdown document, then the markdown document is converted

to HTML, and then the HTML file is what

you view in the, in the Fisa final product.

Generally speaking, you do not want to edit

or mess with any of those secondary documents.

You don't want to edit the markdown document or

the HTML document because those are automatically generated.

If you edit them, and then you reprocess the doc, the original document, all

of your changes will be overridden in the HTML file or in the Markdown file.

So only edit the RMarkdown file that's been,

that contains the original text and the original Rcode.

So here's another example, I've got my first knitr document and I have

my name there as the author, and I've started a section called the introduction,

and say here's some text, here's a code chunk, and I, and I

named the code chunk simulation, and notice

I added this option echo equals false.

And so that means, that indicates the knitr that I don't want the code for this

code chunk to be echoed in the output document, I just want the result to show.

So here I got my level 1 heading indicated by the

single hash, the level 2 heading in, indicated by the two hashes.

I say don't echo the code.

So now this is the output document, again, you could see that

it looks like the original document that I had in the previous example.

But I don't see the code, I'd only see the output, which is the mean.

You could also hide the results, too, if you wanted too.

In this case it doesn't make too much sense, but you could

do it by adding the results, excuse me, the results equals hide option.

And then when you see the output document, you see that all

the text is there but there's no code, and there's no results.

One of the nice things about knitr document

is that you can write sentences and text.

And if you need to fill in a number or a statistic or some sort

of computation in the middle of sentence you

can just add that directly into a sentence.

So for example here I have a code chunk which computes the time

and using the Sys.time() function and then I also generate a random number, okay?

And so, now I can write a sentence.

So that code check, I don't echo the codes, I

don't want people to see that code, what I want people

to see is the sentence immediately afterwards, says the current

time is, and then I'll just plug in the current time.

And then the next sentence is my favorite random number

is, and then I'll just plug in a random number.

So here, you see the output is, the current time

is Wednesday September 4th and there's the time and 2013.

And you can see my favorite random number is 1.1829.

So that's how you can plug in computed variables,

or computed results in the middle, inline with text.

Another important aspect of any report probably is going to be some graphics.

So you want to make plots of data or

of, or maybe, or, or any sort of visualization.

So you can easily add this into an inner document.

So here I've simulated some data.

And then in, in the first arc, code chunk I simulated some data.

And in the second code chunk, I make a plot.

So I use the par function to set the, kind of, the canvas with the margins.

And the las option, then I call plot to make the, make the to generate the plot.

And you notice I added an option here which adjusts the figurehead.

And so actually fig.height equals four which makes

a little bit of squished more than usual.

So it would a be a rectangular type of plot rather than a square plot.

And so what knitr produces is a little bit interesting.

It creates an HTML file through the processing.

So you can see that if you, if you're

familiar HTML you can see this is basic HTML.

I've cut off a lot of stuff from the top, so I've just put the body in.

And this is basic HTML.

I, you can see that down here at the bottom, you'll see there's an image tag.

But rather than point to an actual file that's the image,

it actually embeds the entire image into these HTML files itself.

So you can see this is a Base64 encoded,

image and so actually this HTML file stands alone, it

doesn't, it doesn't depend on an external image files,

which is not necessarily the most efficient format but it's

actually quit useful because you can take this HTML

file, give it to someone else, and they don't have

to worry about whether they have, everything's, about whether

they have these other files or not because everything's embedded.

So here's the result as you can see I

echoed the Rcode I had the result of the plot.

And you see it's kind of this rectangular shape, and that's my scatter plot.

So another handy thing you can do in a knitr document is make a table.

And often you want to make a table that summarizes

the result of some calculation, like a regression model.

So, here I, I have a code chunk with load, with load some data.

And I fiddled in your regression model to it.

Here I'm modelling ozone concentrations, as a

function of wind, temperature, and solar radiation.

And I want to summarize the output from this

Linear Regression model in a, kind of, pretty format.

So, not just your standard, kind of, auto model format.

So I'm going to use the xtable package, which is available on Crane.

You have to install it separately, it doesn't come with R.

And I'm going to use the xtable function to summarize the fit.

And I'm going to use, and I'll print it to HTML.

And you can see that in the resulting knitr, sorry, HTML output, after I

process it with knitr, I get the code that I ran for the regression model.

And I get the, it echoes the code, but at the very bottom here I can see that

I got a table of regression coefficient, standard errors,

T values and P values, which is formatted in HTML.

And, so it, it'll be, kind of in

a prettier format, rather than standard text mode.

So sometimes, you want to set options for the entire document.

So, these are called Global options.

And, you want these options to apply to every single code chunk.

So, maybe if you never want to echo the code, for

example, you want every code chuck to say, eco equals fall.

So it can be a bit of a pain to have to specify every

single code chunk, so you might want to

set a global option for something like that.

And so, the way you can do that is you can create

a separate code chunk in the, in the very beginning of the document.

Here I've called it set options.

And you can modify, there's a special object called opts

underscore chunk, which has a function built into it called set.

And if you call that set function, you can set the global option.

So here I set echo equals to false and I've set the results equal to hide.

So by default, every code chunk in this document will not echo

the result the, sorry, not echo the code and will hide the results.

So here I've got some code chunks here, I simulate some data, I make a plot.

But when I'm, when you see, when I make the,

when I see the HTML output I, I I'm sorry.

Let me go back.

I see in the first of all in the first code chunk I override

the default for the, for the echo so I want to echo equals true.

And then for the second, code chunk, I don't want to echo the plotting code.

And so I, I let the default occur.

So here I, the, the code that generates the data is

echoed but the code that creates the plot is not echoed.

So, some of the common options that are, that

are good to remember, of course you can always go

back to the documentation on the knitr website, so

the results can be specified as either hide or asis.

Asis means don't, don't postprocess the results, just show the,

kind of the, the raw results from whatever the output is.

Echo can be true or false.

So if you want to echo the code or not.

For figures usually you might want to modify the height and the width to be

a specific type of size of fig.height and fig.width are common options to set there.

So one special case is when you have a code

chunk and the computation takes a really long time to run.

Now every time you process the knitr document to look

at the HTML output, you're going to have to run this computation.

So one strategy to avoid having to sit around and wait a long time, is to

cache the computation and so caching computations can

be very useful for kind of complex code chunks.

And so the idea is that for each

code chunk, there's a special option called cache.

You can set cache equal to TRUE and what will happen is that for that

code chunk, the first time you process the document, it will have to run the code.

But in order to figure out what the output is.

But then what it will do is it will store that output on

the disk and so that the next time you process the document, as

long as nothing has changed, it will load the results from the disk,

which is typically a lot faster than having to redo all the computation scans.

So this, this will save you a lot of time if you're doing this over and over again.

So if the data or the code or something changes,

then you will, you will have to rerun the code again.

There's really no getting around that.

And so the caching will, will kind of update to represent the new results.

If you're, if one code chunk depends on another code chunk that

was previous to it, this dependency is not explicitly tracked by default.

And so you'll have to make sure that if a

previous code chunk has a dramatic change to it and the

further and the downstream code chunk depends on it, then you

rerun everything so that the cache doesn't get mixed into it.

Chunks with significant side effects may not be cacheable.

So if a, if the code in a chunk has some effect that

kind of outside the document then it may not be possible to cache that.

So just to summarize, I talk about knitr in this

lecture, which is a useful tool for doing literate statistical programming.

It can be used, that will allow you to put

text, code, and data and output all into one document.

It uses a mark down formatting language, which

is kind of very simple and easy to learn.

And, and it, and it creates HTML documents that you can view in any web browser.

So I think it's a powerful tool for integrating text code.

I mentioned, as I mentioned before, there's pros and cons.

It's not a perfect tool.

But it may be useful to learn and it's for generating [UNKNOWN] reports.

Ознакомьтесь с нашим каталогом

Присоединяйтесь бесплатно и получайте персонализированные рекомендации, обновления и предложения.

Coursera делает лучшее в мире образование доступным каждому, предлагая онлайн-курсы от ведущих университетов и организаций.

© Coursera Inc., 2018 Все права защищены.

Загрузить из App Store

Загрузить в Google Play