재현가능한 과학적 분석을 위한 R: Reference

Key Points

R과 RStudio 소개	RStudio를 사용해서 R 프로그램을 작성하고 실행한다. R에는 수학연산자와 수학함수가 갖춰있다. <-을 사용해서 값(value)을 변수에 할당한다. ls()를 사용해서 프로그램에 사용된 변수를 출력한다. rm()을 사용해서 프로그램에 사용된 객체를 삭제한다. install.packages()을 사용해서 팩키지(라이브러리)를 설치한다.
RStudio와 함께하는 프로젝트 관리	RStudio를 사용해서 일관된 방식으로 프로젝트를 생성하고 관리한다. 원데이터는 읽기 전용으로 처리한다. 자동생성된 출력물은 사용후 버릴 수 있는 것으로 처리한다. 함수 정의와 응용을 분리한다.
도움 청하기	help() 함수를 사용해서 온라인 도움을 얻는다.
자료구조(Data Structures)	read.csv 함수를 사용해서 R에 표형태 데이터를 불러온다. R의 기본자료형은 실수형, 정수형, 복소수형, 논리형, 문자형이다 R에서 인자를 사용해서 범주를 표현한다.
데이터프레임 탐색하기	cbind() 함수를 사용해서 데이터프레임에 칼럼을 추가한다. rbind() 함수를 사용해서 데이터프레임에 행을 추가한다. 데이터프레임에서 행을 제거한다. na.omit() 명령문을 사용해서, NA 값을 갖는 행을 데이터프레임에서 제거한다. levels()과 as.character() 함수를 사용해서 요인을 타맥하고 조작한다. str(), nrow(), ncol(), dim(), colnames(), rownames(), head(), typeof() 함수를 사용해서 데이터프레임 구조를 파악한다. read.csv() 함수를 사용해서 CSV 파일을 불러온다. 데이터프레임 length()가 나타내는 것이 무엇인지 이해한다.
데이터 부분집합 추출	R에서 인덱스는 0이 아니라, 1에서 시작. 위치로 개별값을 접근하는데, []을 사용. [low:high]을 사용해서 데이터 슬라이스 작업 수행. Access arbitrary sets of data using [c(...)]을 사용해서 임의 지점 데이터에 접근하는데 사용. 논리 연산자와 논리 벡터를 사용해서 데이터 일부 부분집합으로 추출.
제어 흐름	if 와 else 를 사용하여 선택을 한다. for 루프를 사용하여 연산작업을 반복한다.
ggplot으로 논문 품질 그래프 생성하기	ggplot2를 사용해서 그래프를 생성한다. 그래프(graphics)를 계층으로 생각한다:aesthetics, 기하(geometry), 통계(statistics), 척도변환(scale transformation), 그룹(grouping).
벡터화(Vectorization)	루프 대신에 벡터화 연산을 사용한다.
함수 설명	function()를 사용해서 신규 함수를 R에서 정의한다. 인자를 사용해서 함수로 값을 전달한다. stopifnot() 함수를 사용해서 유연하게 R에서 함수 인자를 검사한다. source() 함수를 사용해서 프로그램에 저장된 함수를 불러온다.
데이터를 저장하기	‘Export’ 버튼을 사용해서 RStudio로부터 그래프를 저장한다. write.table() 함수를 사용해서 표형식 데이터를 저장한다.
plyr로 데이터프레임을 쪼개고 합치기	plyr 팩키지를 사용해서 데이터를 쪼개고, 쪼개진 데이터별로 함수를 적용시키고 나서, 결과를 조합한다.
dplyr 팩키지로 데이터프레임 솜씨있게 조작	데이터프레임을 조작하는데 dplyr 팩키지를 사용한다. select() 함수를 사용해서 데이터프레임에서 칼럼 변수를 고를 수 있다. filter() 함수를 사용해서 행 값에 따라 데이터를 추출한다. group_by()와 summarize() 함수를 사용해서 데이터 요약 작업을 수행한다. mutate() 함수를 사용해서 변수를 새로 생성한다.
tidyr 팩키지로 데이터프레임 솜씨있게 조작	tidyr 팩키지를 사용해서 데이터프레임 레이아웃 배치를 변경시킬 수 있다. gather() 함수를 사용해서 폭넓은 형태를 긴형태로 변경시킨다. spread() 함수를 사용해서 긴형태를 폭넓은 형태로 변경시킨다.
knitr로 보고서 생성하기	R 마크다운으로 작성한 보고서와 R로 작성한 소프트웨어를 섞어쓴다. 포맷을 제어하는데 덩어리 선택옵션(chunk options)으로 지정한다. knitr 팩키지를 사용해서 문서를 PDF를 비롯한 다른 포맷으로 변환시킨다.
좋은 소프트웨어 작성법	프로젝트 폴더를 구조화 시키고, 잘 조직화하고, 깔끔하게 만든다. 방법(how)이 아닌 무엇(what), 이유(Why)를 문서에 남긴다. 프로그램을 짧고 단일 목적을 갖는 함수로 분해한다. 다시 실행가능한 테스트 코드를 작성한다. 여러분이 반복하지 마라. 여러분이 반복하지 마라. 명칭, 들여쓰기, 다른 코딩 스타타일에도 일관성을 유지한다.

Reference

Introduction to R and RStudio

• Use the escape key to cancel incomplete commands or running code (Ctrl+C) if you’re using R from the shell.
• Basic arithmetic operations follow standard order of precedence:
  • Brackets: (, )
  • Exponents: ^ or **
  • Divide: /
  • Multiply: *
  • Add: +
  • Subtract: -
• Scientific notation is available, e.g: 2e-3
• Anything to the right of a # is a comment, R will ignore this!
• Functions are denoted by function_name(). Expressions inside the brackets are evaluated before being passed to the function, and functions can be nested.
• Mathematical functions: exp, sin, log, log10, log2 etc.
• Comparison operators: <, <=, >, >=, ==, !=
• Use all.equal to compare numbers!
• <- is the assignment operator. Anything to the right is evaluate, then stored in a variable named to the left.
• ls lists all variables and functions you’ve created
• rm can be used to remove them
• When assigning values to function arguments, you must use =.

Project management with RStudio

• To create a new project, go to File -> New Project
• Install the packrat package to create self-contained projects
• install.packages to install packages from CRAN
• library to load a package into R
• packrat::status to check whether all packages referenced in your scripts have been installed.

Seeking help

• To access help for a function type ?function_name or help(function_name)
• Use quotes for special operators e.g. ?"+"
• Use fuzzy search if you can’t remember a name ‘??search_term’
• CRAN task views are a good starting point.
• Stack Overflow is a good place to get help with your code.
  • ?dput will dump data you are working from so others can load it easily.
  • sessionInfo() will give details of your setup that others may need for debugging.

Data structures

Individual values in R must be one of 5 data types, multiple values can be grouped in data structures.

Data types

• typeof(object) gives information about an items data type.
• There are 5 main data types:
  • ?numeric real (decimal) numbers
  • ?integer whole numbers only
  • ?character text
  • ?complex complex numbers
  • ?logical TRUE or FALSE values

  Special types:

  • ?NA missing values
  • ?NaN “not a number” for undefined values (e.g. 0/0).
  • ?Inf, -Inf infinity.
  • ?NULL a data structure that doesn’t exist

  NA can occur in any atomic vector. NaN, and Inf can only occur in complex, integer or numeric type vectors. Atomic vectors are the building blocks for all other data structures. A NULL value will occur in place of an entire data structure (but can occur as list elements).

Basic data structures in R:

• atomic ?vector (can only contain one type)
• ?list (containers for other objects)
• ?data.frame two dimensional objects whose columns can contain different types of data
• ?matrix two dimensional objects that can contain only one type of data.
• ?factor vectors that contain predefined categorical data.
• ?array multi-dimensional objects that can only contain one type of data

Remember that matrices are really atomic vectors underneath the hood, and that data.frames are really lists underneath the hood (this explains some of the weirder behaviour of R).

Vectors

• ?vector() All items in a vector must be the same type.
• Items can be converted from one type to another using coercion.
• The concatenate function ‘c()’ will append items to a vector.
• seq(from=0, to=1, by=1) will create a sequence of numbers.
• Items in a vector can be named using the names() function.

Factors

• ?factor() Factors are a data structure designed to store categorical data.
• levels() shows the valid values that can be stored in a vector of type factor.

Lists

• ?list() Lists are a data structure designed to store data of different types.

Matrices

• ?matrix() Matrices are a data structure designed to store 2-dimensional data.

Data Frames

• ?data.frame is a key data structure. It is a list of vectors.
• cbind() will add a column (vector) to a data.frame.
• rbind() will add a row (list) to a data.frame.

Useful functions for querying data structures:

• ?str structure, prints out a summary of the whole data structure
• ?typeof tells you the type inside an atomic vector
• ?class what is the data structure?
• ?head print the first n elements (rows for two-dimensional objects)
• ?tail print the last n elements (rows for two-dimensional objects)
• ?rownames, ?colnames, ?dimnames retrieve or modify the row names and column names of an object.
• ?names retrieve or modify the names of an atomic vector or list (or columns of a data.frame).
• ?length get the number of elements in an atomic vector
• ?nrow, ?ncol, ?dim get the dimensions of a n-dimensional object (Won’t work on atomic vectors or lists).

Exploring Data Frames

• read.csv to read in data in a regular structure
  • sep argument to specify the separator
    • ”,” for comma separated
    • “\t” for tab separated
  • Other arguments:
    • header=TRUE if there is a header row

Subsetting data

• Elements can be accessed by:
  • Index
  • Name
  • Logical vectors
• [ single square brackets:
  • extract single elements or subset vectors
  • e.g.x[1] extracts the first item from vector x.
  • extract single elements of a list. The returned value will be another list().
  • extract columns from a data.frame
• [ with two arguments to:
  • extract rows and/or columns of
    • matrices
    • data.frames
    • e.g. x[1,2] will extract the value in row 1, column 2.
    • e.g. x[2,:] will extract the entire second column of values.
• [[ double square brackets to extract items from lists.
• $ to access columns or list elements by name
• negative indices skip elements

Control flow

• Use if condition to start a conditional statement, else if condition to provide additional tests, and else to provide a default
• The bodies of the branches of conditional statements must be indented.
• Use == to test for equality.
• X && Y is only true if both X and Y are TRUE.
• X || Y is true if either X or Y, or both, are TRUE.
• Zero is considered FALSE; all other numbers are considered TRUE
• Nest loops to operate on multi-dimensional data.

Creating publication quality graphics

• figures can be created with the grammar of graphics:
  • library(ggplot2)
  • ggplot to create the base figure
  • aesthetics specify the data axes, shape, color, and data size
  • geometry functions specify the type of plot, e.g. point, line, density, box
  • geometry functions also add statistical transforms, e.g. geom_smooth
  • scale functions change the mapping from data to aesthetics
  • facet functions stratify the figure into panels
  • aesthetics apply to individual layers, or can be set for the whole plot inside ggplot.
  • theme functions change the overall look of the plot
  • order of layers matters!
  • ggsave to save a figure.

Vectorization

• Most functions and operations apply to each element of a vector
• * applies element-wise to matrices
• %*% for true matrix multiplication
• any() will return TRUE if any element of a vector is TRUE
• all() will return TRUE if all elements of a vector are TRUE

Functions explained

• ?"function"
• Put code whose parameters change frequently in a function, then call it with different parameter values to customize its behavior.
• The last line of a function is returned, or you can use return explicitly
• Any code written in the body of the function will preferably look for variables defined inside the function.
• Document Why, then What, then lastly How (if the code isn’t self explanatory)

Writing data

• write.table to write out objects in regular format
• set quote=FALSE so that text isn’t wrapped in " marks

Split-apply-combine

• Use the xxply family of functions to apply functions to groups within some data.
• the first letter, array , data.frame or list corresponds to the input data
• the second letter denotes the output data structure
• Anonymous functions (those not assigned a name) are used inside the plyr family of functions on groups within data.

Dataframe manipulation with dplyr

• library(dplyr)
• ?select to extract variables by name.
• ?filter return rows with matching conditions.
• ?group_by group data by one of more variables.
• ?summarize summarize multiple values to a single value.
• ?mutate add new variables to a data.frame.
• Combine operations using the ?"%>%" pipe operator.

Dataframe manipulation with tidyr

• library(tidyr)
• ‘?gather’ convert data from wide to long format.
• ‘?spread’ convert data from long to wide format.
• ‘?separate’ split a single value into multiple values.
• ‘?unite’ merge multiple values into a single value.

Producing reports with knitr

• Value of reproducible reports
• Basics of Markdown
• R code chunks
• Chunk options
• Inline R code
• Other output formats

Best practices for writing good code

• Program defensively, i.e., assume that errors are going to arise, and write code to detect them when they do.
• Write tests before writing code in order to help determine exactly what that code is supposed to do.
• Know what code is supposed to do before trying to debug it.
• Make it fail every time.
• Make it fail fast.
• Change one thing at a time, and for a reason.
• Keep track of what you’ve done.
• Be humble

Glossary

argument

A value given to a function or program when it runs. The term is often used interchangeably (and inconsistently) with parameter.

assign

To give a value a name by associating a variable with it.

body

(of a function): the statements that are executed when a function runs.

comment

A remark in a program that is intended to help human readers understand what is going on, but is ignored by the computer. Comments in Python, R, and the Unix shell start with a # character and run to the end of the line; comments in SQL start with --, and other languages have other conventions.

comma-separated values

(CSV) A common textual representation for tables in which the values in each row are separated by commas.

delimiter

A character or characters used to separate individual values, such as the commas between columns in a CSV file.

documentation

Human-language text written to explain what software does, how it works, or how to use it.

floating-point number

A number containing a fractional part and an exponent. See also: integer.

for loop

A loop that is executed once for each value in some kind of set, list, or range. See also: while loop.

index

A subscript that specifies the location of a single value in a collection, such as a single pixel in an image.

integer

A whole number, such as -12343. See also: floating-point number.

library

In R, the directory(ies) where packages are stored.

package

A collection of R functions, data and compiled code in a well-defined format. Packages are stored in a library and loaded using the library() function.

parameter

A variable named in the function’s declaration that is used to hold a value passed into the call. The term is often used interchangeably (and inconsistently) with argument.

return statement

A statement that causes a function to stop executing and return a value to its caller immediately.

sequence

A collection of information that is presented in a specific order.

shape

An array’s dimensions, represented as a vector. For example, a 5×3 array’s shape is (5,3).

string

Short for “character string”, a sequence of zero or more characters.

syntax error

A programming error that occurs when statements are in an order or contain characters not expected by the programming language.

type

The classification of something in a program (for example, the contents of a variable) as a kind of number (e.g. floating-point, integer), string, or something else. In R the command typeof() is used to query a variables type.

while loop

A loop that keeps executing as long as some condition is true. See also: for loop.