Using R

Part of Data analysis · What links here

Using R

Part Data analysis
Chapter 3 of 9 in this part
Previous Descriptive statistics
Next Estimation examples

R is the language used here to hold a data table, summarise it, draw it, and fit the models in the inference chapters. The objects below are enough to reproduce the summaries in Descriptive statistics and to call the procedures behind Estimation examples, Testing examples, and Linear model examples.

R is installed from the Comprehensive R Archive Network. An editor that sends a script to an R session is convenient. The calculations do not depend on which editor is used. Help for a function is help("mean") or ?mean.

Values, vectors, and assignment

A name is bound with <- or =.

xx <- 10
xx

The printed [1] is the position of the first element, not part of the value. Character data use quotes. mean("I like R") fails because the mean is defined for numbers.

c() builds a vector. Arithmetic and most summaries act elementwise or on the whole vector.

x <- c(2.86, 4.16, 6.35, 3.92, 5.96)
mean(x)
sd(x)
median(x)
length(x)

sd is the bias-corrected sample standard deviation, the s of Estimation. seq(1, 10) and rep(0, 5) build regular vectors. 1:10 is the same idea.

A logical comparison such as x > 4 returns a vector of TRUE and FALSE. Summing it counts the true entries, because TRUE is coerced to 1.

Data frames

A data frame is the spreadsheet of Descriptive statistics: one column per variable, one row per unit. Columns can have different types.

alice <- data.frame(
  change = c(17, 22, 21, 16, 6, -2, 27, 15, 16, 20,
             4, 10, 7, -9, 5, 4, 5, 7, 6, 12),
  drink = rep(c("caffeinated", "decaf"), each = 10)
)
str(alice)
head(alice)

alice$change extracts a column. alice[alice$drink == "decaf", ] keeps the rows that match. summary(alice) prints a numeric summary of each numeric column and a count of each label.

A file of comma-separated values is read with read.csv("heart.csv"). The working directory is getwd(), and setwd() changes it. write.csv(alice, "alice.csv", row.names = FALSE) writes a frame back out.

table(alice$drink) is the count table of a categorical column. tapply(alice$change, alice$drink, mean) applies mean inside each group.

Plots and model fits

hist(x) draws a histogram. boxplot(change ~ drink, data = alice) draws one box per group. plot(x, y) is a scatterplot, and pairs(qvars) is the scatterplot matrix. barplot(table(alice$drink)) draws the bars.

A normal quantile is qnorm(0.975). A t quantile on 9 degrees of freedom is qt(0.975, df = 9). pnorm and pt are the corresponding distribution functions. Random draws are rnorm(n, mean, sd) and runif(n).

The fitted line and the analysis of variance are lm and anova. The formulas and the reading of the coefficients are in Linear regression and Analysis of variance, not in the function names.

fit <- lm(change ~ drink, data = alice)
summary(fit)

lm treats the first level of a factor as the baseline. The intercept is the mean of that level. Each further coefficient is a difference from it. That is the design-matrix form in Linear model examples.

A one-sample t procedure is t.test(x, mu = 0). A two-sample procedure is t.test(change ~ drink, data = alice). A proportion test is prop.test. These functions implement the tests on Hypothesis tests. The printed p-value is the one defined there, not a posterior probability. The posterior calculation is in Bayesian hypotheses and model comparison.

Key idea

Put each unit in a row and each variable in a column. mean, sd, table, and boxplot describe the sample. lm and t.test fit the models already defined in the inference chapters.

See also