• R: Statistical programming language
  • Objects belong to different classes (similar to object-oriented programming)
  • Common data classes include: vector (set of numbers or words), matrix or data.frame (similar to a spreadsheet: columns of vectors), list (usually a mixture of vectors, matrices, or other objects–one key feature is that they are not forced to be the same length)
  • Often, issues with analyses or errors are due to using the wrong type of input class for a function.
  • The great strength of R and other open-source languages (e.g. python, Julia, C, Fortran) is that they are free and you can reproduce your analyses by simply re-running your code!

• Version numbers are used to track changes to software, such as bug fixes and updates. For instance, Windows 10 is a different version than Windows XP.

• When you open R or RStudio, the version number pops up in the Console (the area where code is run in RStudio; in R's graphic user interface (GUI) the default screen is just the console).

• You can also run the command R.Version() or R.version.string

• To find out the version of RStudio go to Help \(\rightarrow\) About RStudio

• It's good practice to keep both R and RStudio up to date
  • It can be painful to deal with re-installing R to update it, but there are good resources on how to deal with packages, such as r-bloggers. Note that if you update RStudio, you often have to update the R language version as well.

• A workspace is "your current R working environment and includes any user-defined objects (vectors, matrices, data frames, lists, functions) that you have entered. At the end of an R session, the user can save an image of the current workspace that is automatically reloaded the next time R is started." Quick-R

• For instace, we can see what is in my workspace right now using:

ls()

• We can also see what directory (file location or path) R is running out of.

getwd()

• If you have specific objects (such as analyses or tables, or anything else) that you want to save to re-load later (and avoid re-running certain bits of code that might be slow or memory-intensive), you can use the command save or save.image.
• Example: I have a regression (regression_mod) that I want to save:

save(file="~/Rscripts/regression_mod.RData",regression_mod) 
# Unix (Linux/Mac) style path

• The alternative is to call save.image which would save everything in the current workspace.
  • This is similar to the Quit R Session: Save Workspace Image? prompt, but better:
  • You control where the objects are located (file name)!

save.image(file="C:\Documents\MyWorkspace-5-26-2018.RData") # Windows-style path

• You can then re-load these objects into a new session with the commands:

load(file="C:\Documents\MyWorkspace-5-26-2018.RData")

• "Packages are the fundamental units of reproducible R code. They include reusable R functions, the documentation that describes how to use them, and (often) sample data." (From Hadley Wickham, a prominent R software developer with many helpful resources, emphasis mine.)

• R packages are stored online at the Comprehensive R Archive Network (CRAN), which currently has 12579 packages.

• The command install.packages lets you easily install any of the packages in CRAN.

install.packages("gdata") # a package that specializes in dealing with excel spreadsheets
install.packages("ggplot2") # a popular package for plots that is an alternative to base R

• Note that you only have to install each package once!
  • It is like downloading a file or software: once it is on your machine, it's on your machine!

Let's get on to an example of real-world data.

Here is a simulated (fake!) dataset entitled xishuangbanna_birds_df with bird counts and an ecological covariate (eco_cov).

## # A tibble: 26 x 6
##    site   eco_cov Bulbul Hornbill TitBabbler Babbler
##    <fctr>   <dbl>  <dbl>    <dbl>      <dbl>   <dbl>
##  1 A         7.49   5.00     1.00       3.00    2.00
##  2 B        10.7    7.00     1.00      11.0     1.00
##  3 C         9.61   9.00     1.00       2.00    1.00
##  4 D        14.4    9.00     1.00      12.0     4.00
##  5 E        10.6    9.00     1.00       8.00    1.00
##  6 F        11.6    7.00     1.00       5.00    3.00
##  7 G         7.09   6.00     1.00       6.00    3.00
##  8 H        13.6   12.0      1.00       7.00    1.00
##  9 I         5.87   6.00     1.00       3.00    1.00
## 10 J         8.20   7.00     1.00       5.00    1.00
## # ... with 16 more rows

• For the rest of this short course, we will continue with instruction focused on standard base R syntax (not tidyverse syntax).
• Oftentimes, you will want to import data from a spreadsheet. read.csv is perfect for Comma Separated Value (CSV) spreadsheets

read.csv(file="C:\Documents\FieldData.csv", header=T) 
  # Note that you have to specify the path (directory) where your file is located.
      # In this example, this would be a Windows-style path
  # header: do you have variable names for each column?

• The more general file read-in function is read.table which can take .txt, .csv, etc.

read.table(file="~/Documents/FieldData.csv",header=F)
  # In this example, the file directory is for a Unix (Linux/Mac OS) system
  # header=F means that I did not have names for the columns 
  # (no names for the variables), 
  # so they will be named X1, X2, X3, etc.

• For Excel spreadsheets (*.xls, *.xlsx), the package gdata has a function to read in excel files that avoids common input issues.

install.packages("gdata") # recall that you only need to do this ONCE!
library("gdata") # you must do this every time 
                 # you start a new R session (re-open R)

## There are two ways to use the function read.xls from the gdata package:

    # 1: call it directly
read.xls(file="C:\Documents\FieldData.xlsx", sheet=1) 
# read.xls can only extract from one spreadsheet page at a time, 
# so tell it which page to look at

    # 2: tell R to go and look for the function "read.xls" from package "gdata":
gdata::read.xls(file="C:\Documents\FieldData.xlsx",sheet="Birds") 
# if your sheets have names, you can also refer to them by the names of each tab.

• You can use R to perform addition, subtraction, multiplication (including matrix multiplication), division, and pretty much any mathematical task you desire.
• Let's start by creating a variable x and assigning the value of 5 to it:

x <- 5
x

## [1] 5

x + 5

## [1] 10

x / 5

## [1] 1

x * 5

## [1] 25

x - 5

## [1] 0

• R has many standard functions for descriptive statistics, for example:

mean(xishuangbanna_birds_df$Hornbill)

## [1] 1.038462

sd(xishuangbanna_birds_df$Hornbill) 

## [1] 0.1961161

median(xishuangbanna_birds_df$Hornbill)

## [1] 1

summary(xishuangbanna_birds_df$Hornbill)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   1.000   1.000   1.038   1.000   2.000

• Sometimes, you may have missing values. These are known as nas. To deal with them, you can use na.rm=T: take out any missing values from my data before you do the analysis

# Sample 10 observations from a poisson distribution
spotty_data <- rpois(8, 10)
# Remove the 5th observation and replace with NA
spotty_data[5] <- NA
mean(spotty_data) # uh-oh!

## [1] NA

mean(spotty_data,na.rm=T) # phew!

## [1] 10.57143

• In Base R, you can use the command plot to create scatterplots, boxplots, histograms, barplots, and many other types of plots!
• The package ggplot2 further extends the plotting capability of R

par(mfcol=c(1,3), oma=c(0,0,0,0))
plot(cars,xlab="Speed",ylab="Distance",main="Cars Dataset Scatterplot", pch=19, col="grey60")
boxplot(count ~ spray, data = InsectSprays, col = "lightgray",xlab="Treatment",ylab="Insect Count", main="Insect Sprays Dataset")
hist(islands,xlab=expression(paste("Land area (miles"^2,")")))

• It is straightforward to build regression models (how does y (dependent variable) relate to x (predictors/independent variables)?)
  • More regression examples here
• As an example, let's regress braking distance against car speed using the cars dataset (?cars):

data(cars)
# build linear regression model on full cars dataset
# and store in a variable named linearMod
linearMod <- lm(dist ~ speed, data=cars)  
linearMod # show me linearMod

## 
## Call:
## lm(formula = dist ~ speed, data = cars)
## 
## Coefficients:
## (Intercept)        speed  
##     -17.579        3.932

# What was the model fit?
summary(linearMod)

## 
## Call:
## lm(formula = dist ~ speed, data = cars)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -29.069  -9.525  -2.272   9.215  43.201 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) -17.5791     6.7584  -2.601   0.0123 *  
## speed         3.9324     0.4155   9.464 1.49e-12 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 15.38 on 48 degrees of freedom
## Multiple R-squared:  0.6511, Adjusted R-squared:  0.6438 
## F-statistic: 89.57 on 1 and 48 DF,  p-value: 1.49e-12

• Bayesian approaches can be particularly helpful when you are dealing with data that come from a complex, hierarchical generative process.
• Take bird survey data as an example.
• In bird surveys, we face multiple problems:
  • A species might be present, but we may fail to detect it.
  • Certain species may be harder to detect at some times of the day or year.
  • Habitat quality can affect both the abundance of species and their detectability (e.g. how much are they singing, how active are they)?
• Bayesian models can allow us to specify the set of interactions leading to changes in detectability and abundance.
• There are many helpful resources on Bayesian methods in R:
  • Gelman's Bayesian Data Analysis
  • Statistical Rethinking
  • Kery and Royle

• Markdown is a straightforward language for turning plain text into nicely formatted HTML or PDF, which is a great format for sharing with collaborators.

• R Markdown combines the syntax (language rules) of markdown with embedded R code chunks that are run so their output can be included in the final document.

• R Markdown lets you create documents (you could even write an entire scientifc manuscript in R Markdown!), presentations (this presentation is written in R Markdown), and reports from R.

• R Markdown documents are fully reproducible (they can be automatically regenerated whenever underlying R code or data changes). This is a huge advantage over a static workflow such as Microsoft Word + SPSS. Whenever you want to re-run your analyses, simply alter your code (usually just some small part of your code)–this is much easier than having to go back into something like SPSS or SAS and re-do all of the manual data input and analyses.

Source: http://rmarkdown.rstudio.com/ and https://github.com/mine-cetinkaya-rundel/rworkshop-mem/

In RStudio, you can create R Markdown documents, "knit" it (render the code into a nice format such as HTML or PDF), and examine the source code and the output.

1. File \(\rightarrow\) R Markdown…

2. Enter a title (e.g. "Spatial analysis of predators and prey") and author info

3. Choose Document as file type, and HTML as the output

4. Hit OK

5. Click Knit HTML in the new document, which will prompt you to save your document
   • Naming tip: Do not use spaces
     • Instead, use "CaMeLcAsE" or underscores ("Camel_Case").
   • Viewing tip: Click on the down arrow next to Knit HTML and select View in Pane

• Markdown is a very simple formatting language based on plain text (the standard text format used in emails, at least historically)

• Rather than writing complex code (e.g. HTML), Markdown enables the use of a syntax much more like plain-text email.

Markdown overview

Within an R Markdown file, R Code Chunks can be embedded using the native Markdown syntax for fenced code regions.

Code chunks