There are a lot of questions and answers on the internet. This page lists resources that I have found useful, so I thought you might find them useful too.
Step one: download R to your appropriate system (Windows, Mac, and Linux).
Step two: download RStudio. RStudio has invested a lot of time and money to make using R much, much easier. It’s also free!
Step three: Read R for Data Science. This is a great guide written by Hadley Wickham, a prolific R programmer who has had fantastic positive impact on the landscape of R packages and practices. It introduces the principles of the tidyverse
- a set of R packages taht play really well together. They cover tools to data reading in/out, data manipulation, data visualisation, and tools for performing and summarising models.
If you are having problems with getting R or RStudio, check out this guide, from this guy. It covers installing R and RStudio in a little more detail.
To learn R, you need to learn how to get unstuck with R. This teaches you a really good process to iterate through when going through the process of getting unstuck.
To learn a new function or package, Quick-R, provides nice quick description of functions and other R-related things.
For all my other problems, I usually google the error message, or try my darndest to ask an reasonable question to google that describes my current dilemna, and then look read the appropriate blog post, or StackOVerflow Answer. RSeek is also basically a google search that filters by R related content.
I would also recommend checking out RStudio’s list of resources for learning R, and this blog post, which describes learning R from a social sciences background.
To stay up to date with what other people around the world are doing with R, I recommend checking up on r-bloggers every other day, and checking out the #rstats hashtag on twitter. The R and statistics community on twitter is both excellent and friendly.
Got a basic handle on R and are hankering for more? I recommend these free, online books by Hadley Wickham:
There is also a book, Ramarro, by quantide which seems similar(ish) to Hadley’s books.
R’s S3 classes are this really awesome minimal class of functions that can be super handy in R. They are described nicely in Hadley’s book, but I have also found these to be helpful:
This blog post, which also has such a suave blog layout.
This video by Andrew Robinson. Sides available here. Thanks to http://damjan.vukcevic.net/ for this information.
I have also written a blog post about S3 methods, and have a preprint on arxiv.
ggplot2
If you are going to do a plot in R, it should be in ggplot. It takes about 5 minutes to get the hang of, and once you’ve got it down you can create plots that make sense, behave how you expect, and look fantastic.
ggplot ollows a logical syntax adapted from the book “The Grammar of Graphics”. It makes visualisation make sense. And there are lots of other packages that build upon it to make it more awesome, such as GGally, ggalt, ggExtra, ggforce, gganimate, and ggbeeswarm, to name a few!
Here are some ggplot resources in order of usefulness
Plotly for R, written and maintained by Carson Sievert, is a very powerful and flexible interactive plotting enging in R. It has a fully fledged API for writing interactice graphics in R, as well as a fantastic function that gives the user a lot for free: ggplotly
. You can read more about plotly for R in Carson’s free and online book.
ggvis is another great package written by Hadley Wickham, which builds upon the structure of ggplot but it allows for more interactive, reactive, plot building. Examples can be found here here, and here.
More serious development on ggvis will apparently begin in 2018, as Hadley and his team at RStudio will be spending 2017 to make the everything in the tidyverse work well together. For the moment I would recommend using plotly to do your interactive graphics, although ggvis is still great!
shiny is a really awesome way to enhance your R script, package, or method. Shiny turns these into ‘apps’, that people can interact with.
tidyr::replace_na
.Probably the coolest thing ever. knitr
is this amazing package that allows the user to combine their code and document text, making research easier to reproduce, and it does this while looking slick and classy. The idea is essentially to let the human do the writing, and the computer handle displaying the results, so that reports can be easily constructed, and most importantly, reproduced easily.
Check out some really nice guides here and here, and from the awesome dude who created knitr here.
You can also augment your rmarkdown documents with templates. For example - rticles which is an r package that adds loads of rmarkdown templates. Currently, there are templates for the R Journal, the UseR Conference, Journal of Statistical Software, PLoS Computational Biology, and more!
If you want to learn statistics using R, check out this website containing 15 hours of an applied R statistics course from Stanford. They also have an excellent (and free!) book.
I use decision trees a lot in R, and I even wrote a little package that helps take care of some common tasks in interrogating decision trees. Here are a list of resources that I recommend using to learn about them:
This book from James et al - chapter 8 specifically refers to decision trees. They’ve also made the book free! Also their videos on decision trees are very useful. You can find a comprehensive list of all their videos and material at this website
This book chapter from the Handbook of Statistics is broad and general.
This page helps explain regression trees. Their gif demonstrating how decision trees choose splitting values is also really helpful.
This video on introduction to boosting trees for regression and classification by statsoft.
Spatial data analysis can be really different to anything else that you’ve done in R. Well, it was for me. Fortunately, recent awesome progress has been made on the simple features R package, officially supported by the RConsortium, and authored by Edzer Pebesma. The format of simple features is to adopt a standard dataframe format, where every row is a spatial feature, and the spatial features are described in a geometry list column. This is really fantastic, because it means that (for the most part), working with spatial data is very similar to working with regular dataframes, which is the bread and butter of analysis and data wrangling in R.
In particular, simple features is designed to play nicely with the tidyverse, and accordingly plays well with ggplot2, dplyr, purrr, and so on. It’s amazing. Anyway, here is a list of resources on using the simple features format:
The R Spatial Blog is a great way to stay updated with the latest changes in simple features.
A blog post by Matt Strimas-Mackey on how to use simple features with dplyr, tidyr, and ggplot2.
For more thoughts on R for spatial data analysis:
For interactive visualisation of spatial data, I really like RStudio’s leaflet, which is a port of the excellent JavaScript leaflet library is my go to place.
ggmap, is also great, as it produces static maps.
STATA do a great job of explaining multilevel and hierarchical models on their blog. I found these two blogs and video really helpful:
Just as it is important to have strong data visualisation skills, it is important to understand what makes a good looking document, poster, business card, and whatnot. To this end, you should read typography in ten minutes, and the summary of key rules of typography. One day I will purchase some fonts to pay him back.