# ----------------------------------------------------------------------------- # Part 0. Getting around ------------------------------------------------------- # ----------------------------------------------------------------------------- # RStudio panes: Source (scripts), Console (runs code), Environment (objects), # Files / Plots / Help. # Anything after "#" is a comment and is not run. ?mean # open the help page for a function example(mean) # run the examples from the help page getwd() # which folder R is currently working in # ----------------------------------------------------------------------------- # Part 1. Value and variable -------------------------------------------------- # ----------------------------------------------------------------------------- # Arithmetic 2 + 3 10 / 4 2 ^ 10 sqrt(81) log(100, base = 10) # functions take named arguments # Variable x <- 5 # "<-" assigns a value to a name (read it as "gets") y <- x * 2 y # Basic data types class(3.14) # "numeric" class("hello") # "character" class(TRUE) # "logical" class(2L) # "integer" (L makes it a whole number) # Question: What does 5 == 5.0 return? What about "5" == 5? # ----------------------------------------------------------------------------- # Part 2. Vectors and logic --------------------------------------------------- # ----------------------------------------------------------------------------- scores <- c(72, 85, 90, 64, 78, 95) # c() combines values into a vector length(scores) mean(scores); median(scores); sd(scores) # Vectorisation scores + 5 scores / 100 # Indexing scores[1] # the first element (R counts from 1) scores[2:4] # a slice scores[-1] # everything except the first element # Logical filtering scores > 80 # a TRUE/FALSE vector scores[scores > 80] # keep only the values that are TRUE sum(scores > 80) # TRUE counts as 1, so this counts the values mean(scores > 80) # proportion above 80 # Handy sequences 1:10 seq(0, 1, by = 0.25) rep("A", 3) # Question: How would you get the scores between 70 and 90 (inclusive)? # ----------------------------------------------------------------------------- # Part 3. Functions and simple control flow ----------------------------------- # ----------------------------------------------------------------------------- to_z <- function(v) { (v - mean(v)) / sd(v) # the last line is returned automatically } round(to_z(scores), 2) grade <- function(s) { if (s >= 90) { "A" } else if (s >= 80) { "B" } else { "C or below" } } grade(85) # For loop for (s in scores) { cat(s, "->", grade(s), "\n") } # A vectorised equivalence sapply(scores, grade) # apply grade() to every element # ----------------------------------------------------------------------------- # Part 4. Data frames --------------------------------------------------------- # ----------------------------------------------------------------------------- # mtcars: 32 cars from a 1974 magazine (mpg = miles per gallon, # cyl = number of cylinders, wt = weight in 1000 lb, am = 0 automatic / 1 manual) head(mtcars) # first six rows str(mtcars) # structure: size, column names, types summary(mtcars) # quick summary of every column dim(mtcars); nrow(mtcars); names(mtcars) # Picking columns and rows mtcars$mpg # one column, as a vector mtcars[1:3, c("mpg", "cyl", "wt")] # data[rows, columns] mtcars[mtcars$mpg > 25, ] # rows that meet a condition subset(mtcars, cyl == 4 & am == 1, select = c(mpg, wt)) # easier to read # Adding a new column mtcars2 <- mtcars # work on a copy mtcars2$kpl <- mtcars2$mpg * 0.425 # kilometres per litre mtcars2$transmission <- ifelse(mtcars2$am == 1, "manual", "automatic") head(mtcars2[, c("mpg", "kpl", "transmission")]) # Sorting head(mtcars2[order(mtcars2$mpg, decreasing = TRUE), c("mpg", "cyl")]) # Making your own data frame students <- data.frame( name = c("Ana", "Ben", "Cai", "Dee"), year = c(3, 4, 3, 4), score = c(88, 92, 79, 85) ) students # ----------------------------------------------------------------------------- # Part 5. Missing values (N/A) ------------------------------------------------ # ----------------------------------------------------------------------------- # airquality: daily air readings in New York, May to September 1973 summary(airquality) # note the NA's under Ozone and Solar.R mean(airquality$Ozone) # NA, because one missing value spreads mean(airquality$Ozone, na.rm = TRUE) # drop the NAs first sum(is.na(airquality$Ozone)) # how many are missing? colSums(is.na(airquality)) # missing values in each column complete <- na.omit(airquality) # keep rows with no missing values nrow(airquality); nrow(complete) # ASK: Is dropping rows always safe for statistics? # ----------------------------------------------------------------------------- # Part 6. Group summaries ----------------------------------------------------- # ----------------------------------------------------------------------------- table(mtcars$cyl) # counts table(mtcars$cyl, mtcars$am) # two-way table tapply(mtcars$mpg, mtcars$cyl, mean) # mean mpg for each cyl group aggregate(mpg ~ cyl + am, data = mtcars, FUN = mean) # formula interface # ----------------------------------------------------------------------------- # Part 7. Base R plots -------------------------------------------------------- # ----------------------------------------------------------------------------- hist(mtcars$mpg, main = "Fuel efficiency", xlab = "Miles per gallon", col = "lightblue", breaks = 8) boxplot(mpg ~ cyl, data = mtcars, main = "MPG by number of cylinders", xlab = "Cylinders", ylab = "MPG", col = c("#9ecae1", "#fdae6b", "#a1d99b")) plot(mpg ~ wt, data = mtcars, pch = 19, col = "steelblue", main = "Heavier cars use more fuel", xlab = "Weight (1000 lb)", ylab = "MPG") abline(lm(mpg ~ wt, data = mtcars), col = "red", lwd = 2) # fitted line # ----------------------------------------------------------------------------- # Part 8. A first look at statistics ------------------------------------------ # ----------------------------------------------------------------------------- cor(mtcars$mpg, mtcars$wt) # correlation # Do manual cars get better mileage than automatic ones? (Welch two-sample t-test) t.test(mpg ~ am, data = mtcars) # Linear regression fit <- lm(mpg ~ wt + hp, data = mtcars) summary(fit) # coefficients, R-squared, p-values coef(fit) predict(fit, newdata = data.frame(wt = 3, hp = 150)) # a prediction for a new car