class: center, middle, inverse, title-slide # Good Practices in R ## A tutorial for Pasadena RLadies ### Zhi Yang ### 2019/1/9 (updated: 2019-03-05) --- # This talk is `not` for those who... - haven't used RStudio -- - haven't used R for three months or worked on more than `3` projects -- - expect to see a systematic review like a book or a workshop --- # This talk is actually ... -- - from my daily regimen of programming in R -- - about **good** practices from my own view, aka, `biased` -- - my experiences of gradually adopting new programming habits --- # Who am I and what do I use R for? -- - A Ph.D. student in biostatistics from USC -- - I use R for > 5hrs/per day and my dissertation is dependent on R -- - I self-taught R in 2013 from a reproducible project in R -- - I've done R Markdown, R packages, Shiny apps, high-performance computing, data visualization, Bayesian statistics, machine learning, version control with Git and etc in R. -- - Some other non-academic projects with R: ranking movies by ratings from my MoviePass membership, web scraping my favoriate brand of clothes, ... -- - `NEW YEAR RESOLUTION`: an intermediate user who would like to become an advanced learner in 2019 🤓 --- ## I know running things in RStudio, now what? -- #### Start a new analysis - Open RStudio and create a new R script - Too many R scripts in a folder -- `-> RStudio project` -- #### Retrieve an old analysis - What is the name of that file again? test.R, test2.R, final.R, or project.R? -- `-> Naming files` -- #### Run an analysis in RStudio - Click run bottom for a thousand times - Use `=` instead of ` <- ` without any indentation or surrounding spaces - overwelmed by seeing too much code -- `-> Customize RStudio`, `Code style` --- class: inverse, center, middle # Get Started --- # Download RStudio https://www.rstudio.com/products/rstudio/download/  --- # Customize panes  --- # Customize panes  --- # Don't go by default  --- # Don't go by default  --- # Always start with a project  --- # Always start with a project  --- # Use relative paths ###NO! ```r pathName <- "/Users/Zhi/Dropbox/PRL_GoodPractice/files/dataset-2019-01.csv" ``` -- ###YES! ```r pathName <- "data/dataset-2019-01.csv" ``` --- # Google's R style guide What [Google](https://google.github.io/styleguide/Rguide.xml) suggests about the style of programming in R  --- # Name files Recommended by @Emil_Hvitfeldt ###NO! ```r - test.R, test2.R - final.pdf - project.rmd - analysis20180501.r ``` -- ###YES! ```r - 2019-01-09_pm511a_finalproject.r - 01_pm511a_finalproject.r - analysis_pm511a_finalproject.r ``` --- # Name identifiers ###NO! ```r - "Variables": avg_Clicks, - "Functions": calculate_avg_clicks, calculateAvgClicks ``` -- Don't use underscores ( _ ) or hyphens ( - ) in identifiers -- ###YES! ```r - "Variables": avgClicks, avg.clicks - "Functions": CalculateAvgClicks ``` -- Generally, variable names should be nouns and function names should be verbs. --- # Format codes (be nice to yourself 👁👁) -- ```r genRandom=function(n){rnorm(n,10,10)} plot(2^seq(2,10,0.5),sapply(sapply(2^seq(2,10,0.5),genRandom),mean)) ``` -- - Do use ` <- ` instead of `=` - Do pipe `%>%` instead of `((()))` - Do use spaces - Do wrap long lines - Do indent lines with two spaces -- Instead, we can write codes like ```r libary(magrittr) xVar <- 2^seq(2,10,0.5) genRandom <- function(n) { rnorm(n, mean = 10, sd = 10) } sapply(xVar, genRandom) %>% sapply(mean) %>% plot(xVar, .) ``` --- # Still not interested? The `formatR` package, by Yihui Xie, makes it easier to clean up poorly formatted code. It can’t do everything, but it can quickly get your code from terrible to pretty good. -- Hadley Wickham ```r library(formatR) genRandom=function(n){rnorm(n,10,10)} ``` -- Copy the above code to clipboard and type `tidy_source(width.cutoff = 50)` ```r genRandom = function(n) { rnorm(n, 10, 10) } ``` --- # Comment your codes ```r libary(magrittr) *# explore the number of simulations in a wide rang------------------------- xVar <- 2^seq(2,10,0.5) *# write a function to simulate n numbers of standard normally distribution R.V.s genRandom <- function(n) { rnorm(n) } *# create a plot with mean against a range of number of simulations sapply(xVar, genRandom) %>% sapply(mean) %>% plot(xVar, .) ``` Have all the library calls and hard-coded variables on the topic of the script -- ### Be nice to your future self three months from now 🤗 --- # Write loops and functions `apply`, `sapply`, `lapply`, `replicate` help to declutter codes ###NO! ```r rnorm(10, mean = 0, sd = 1) rnorm(1e2, mean = 0, sd = 1) rnorm(1e3, mean = 0, sd = 1) rnorm(1e4, mean = 0, sd = 1) ``` -- ###YES! ```r n <- 10^(1:4) n %>% sapply(function(x) rnorm(x, mean = 0, sd = 1)) ``` Have a file with functions in it and source it after loading libraries --- # Organize files ```r Project ├── code ├── data ├── fig ├── result ``` input and output ```r read.csv(data/dataset.csv) ggsave("fig/mtcars.png") ``` --- # Create shortcuts - `Ctrl+X/C/V/Z/Y/S ->` Cut/Copy/Paste/Undo/Redo/Save -- - `Ctrl+Enter ->` Run the current code -- - `Ctrl+Shift+K ->` Knit Markdown -- - `Alt + - ->` <- -- - `Ctrl+Shift+M ->` %>% -- - `Ctrl+Shift+F10 ->` Restart R --- # Have troubles? - https://stackoverflow.com/ - https://support.rstudio.com - https://twitter.com/ `#rstat#` --- # Ask questions reproducibly  --- class: center, middle # More resources Beyond Basic R: https://owi.usgs.gov/blog/intro-best-practices/ Emil Hvitfeldt's talk : https://github.com/EmilHvitfeldt/oRganized-talk --- class: center, middle # Thanks! and Keep in touch <br> ### @zhiiiyang ### zyang895@gmail.com ### Slides created via the R package [**xaringan**](https://github.com/yihui/xaringan) Template created by [Alison Hill](https://twitter.com/apreshill)