"Education is not the filling of a pale, but the lighting of a fire"

W.B. Yeats

• The course will be three days of relaxed learning
• Day 1 will be entirely focused on getting familiar with R
• The day will begin with getting everything installed and a short presentation providing an overview of R (10:30 - 12:30)
  • Capabilities + syntax
• Then we will do a short practical to hone our skills (13:30 - 15:30)
• For the final session we can have an interactive session where you can provide challenges and we can work through the solutions together (maybe with your own data)

• A programming language
  • Computational problems
  • Academia
  • Industry
• A statistical package
  • Basic stats (t.test, anova, regression)
  • Advanced stats (random forests, clustering, machine learning)
• A scripting environment
  • Building complex analysis pipelines
  • Bioinformatics

• It is used widely in academia and industry

• It is currently one of the most popular analytical software used among professional data scientists (academic + commercial)

• Positives
  • Fully functional statistical programming environment
  • Open source
  • Cross-platform (Windows/Linux/Mac)
  • Excellent graphing capabilities
  • Thousands of free extension packages
  • Large online community for support
  • Reproducible research
  • High level syntax (relatively easy to learn and code)

• Negatives
  • Steep learning curve (Lots of help available)
  • Minimal GUI capabilities
  • Analysing large data sets can be troublesome (not impossible)
  • Scripts cannot be compiled into stand alone .exe programs (web apps?)
  • Interpreted language (slow compared to compiled C++ etc.)
  • Thousands of (free) extension packages

• Simple arithmetic

    5 * 5

  [1] 25

• Advanced calculator \[I_n(Q;J=j) = -p_{j}log_{e}p_{j} + \sum_{i=1}^K\frac{p_{ij}}{K}log_{e}p_{ij}\]

  In[j] <- (-(p[j])*log(p[j])) + sum((p[i, j]/k)*log(p[i, j]))

• Data analysis
  • General statistical tests (t-tests/anova/glm etc.)
  • Specialist analysis (abc/clustering/networks etc.)
• Data visualisation
  • Basic plotting

• More complex plotting

• Design web applications using the shiny package
  • diveRsity-online (diveRsity package app for popgen analysis)
  • divMigrate-online (web app for directional gene flow analysis)
  • k-means cluster (demo of k-mean clustering of Isis data)
  • Global sea surface temperature

• Choice of integrated development environment (IDE) is essential for convenient and efficient projects (e.g. analysis for paper x)

• They also provide some niceties such as syntax highlighting, code completion, and code diagnostics (RStudio only)

• There are a number of choices available
  • Eclipse
  • Tinn-R
  • Emacs (ESS)
  • RStudio IDE

• Why RStudio?
  • Well supported
  • The most actively developed IDE
  • Some of the best R developers in the world involved (Hadley Wickham, JJ Alaire)
  • Much more than R
    • \(\LaTeX\) (Write Word/PDF/HTML papers in RStudio)
    • markdown (Write reports, papers or presentations, e.g. this presentation)
  • Version control (git + SVN)
  • Projects
  • Lots of new features with each release

• Although RStudio has minimised the practical importance of this, it is still essential to understand it

• The working directory is the place where R looks for files that you ask to read, and the place where results will be written

• The working directory can be determined using:

  getwd()

• To manually specify a new working directory, use:

  setwd("path/to/new/folder")

• Although the best place to find help when using R is online, there is a more formal build-in help system

• All functions have a dedicated help file that can be accessed as follows:

  ?function_name
  # or
  help(function_name)

• Providing you have got the name correct, this will open a help file

• We will see some examples of these files later

• There are many useful websites dedicated to using R
  • http://www.ats.ucla.edu/stat/r/
  • http://www.rseek.org/
  • http://stackoverflow.com/questions/tagged/r
  • Twitter #rstats #rlang
  • http://www.r-bloggers.com/
  • http://en.wikibooks.org/wiki/R_Programming
  • http://www.statmethods.net/
  • http://adv-r.had.co.nz/

• Hundreds of books for all kinds of research using R

• A list of almost every book ever written for R
  • This site also appears to have links to pdf version of many of the books. I am not clear on the legality of this so exercise caution.
• Each other!
  • One of the major benefits of R is the active and enthusiastic community
  • There are some 'bad eggs' who make learning more difficult than it should be, but the majority understand the challenge.

• The true power of R comes from the 1000s of additional packages available

• These can be installed using:

  install.packages("package_name")

• A growing number of packages are available on github
  • Difficult to find the gems
  • May have poor documentation and bugs

library(devtools)
install_github("username/repo")

• R can be used as a simple calculator

5 + 5

[1] 10

• And a less simple calculator

log10(10)

[1] 1

• Common mathematical operators

^ or **  powers: 2^10 == 2**10
* and /  Mutliplication and division
+ and -  Addition and subtraction
%/%      Integer division
%*%      Conformable matrix multiplication
etc. etc.

• Common build in mathematical functions

  sin()|cos()|tan()|log()|log10()
  sqrt()|sum()|floor()|ceiling()
  round()|abs()|acos()|atan()|factorial()

• Example

sum(1,3)

[1] 4

sum(1,3) == 1 + 3

[1] TRUE

• All variables in R will have a type
• This information is important for a number of reasons
  • A numeric type can't be multiplied by a character type

    5 * "a"

  Error in 5 * "a": non-numeric argument to binary operator

• Certain data structures can only contain as single data type
  • vector
  • matrix
• Others allow a mix of types
  • lists
  • dataframe

• There are simple and complex data structures in R

  • Simple: all values have the same type (e.g. all numeric)

    matrix
    vector
    array

  • Complex: values can have multiple/complex types

    factor()
    list()
    data.frame()

• Creating a numeric vector named x

x <- c(1, 2, 3)
x

[1] 1 2 3

• Creating a character vector named y

y <- c("Kevin", "Keenan")
y

[1] "Kevin"  "Keenan"

• What happens if we try to create a vector z from x and y?

z <- c(x, y)
z

[1] "1"      "2"      "3"      "Kevin"  "Keenan"

• This is know as coercion, and is important to be aware of

• What if we really did want a variable with both numeric and character values

z <- list(numbers = x, strings = y)
z

$numbers
[1] 1 2 3

$strings
[1] "Kevin"  "Keenan"

• This feature is extremely useful for statistical computing

• Although population genetics analysis is slightly different from standard analyses in R, a number of 'standard' skills will come in handy for:
  • Data visualisation
  • Taking results beyond the basics
• Most data analysis (and manipulation) tasks are carried out on the dataframe. However, other data structure like the matrix etc. may come in useful.

• Let's see what we mean by vector of different types

str(dat)

'data.frame':   5 obs. of  3 variables:
 $ sex: Factor w/ 2 levels "F","M": 2 2 1 2 2
 $ hgt: num  75.2 76.1 75.8 75.3 75.2
 $ col: Factor w/ 4 levels "black","blond",..: 2 1 4 3 4

• A special vector in R, where elements take on one of a set of values know as levels

dat$sex

[1] M M F M M
Levels: F M

• Factors are very useful for grouping variables in statistical analyses and plotting

tapply(X = dat$hgt, INDEX = dat$sex, FUN = mean)

     F      M 
75.840 75.465 

• As R is a programming language, it is also possible to write functions (programs/modules/routines)

fun_name <- function(arguments){
  work
  done
  return(results)
}

• We will get the chance to write our own functions later

• Useful for applying an expression many times
• These are some of the most useful functions in R
  • apply
  • lapply
  • sapply
  • mapply
  • by
  • rapply
  • eapply
• Differ in both input and output structure
• Supersede the need for for loops