The special variable .Last.value holds the value of the most recent expression at the command line.

veryLongRunningFunction()   # Oops! Forgot to save result
x <- .Last.value            # Save it now

Careful: .Last.value is overwritten with every new expression you type, so use it immediately!

x <- 3     # Assign 3 to x
3 -> x     # Does the same thing

# Handy for capturing results
veryLongExpression +
  anotherVeryLongExpression *
  yetAnotherVeryLongExpression -> x

y %>% f
y %>% f %>% g
y %>% f %>% g %>% h -> x

Very handy at the command line. Can’t recommend it for scripting: too odd.

Double brackets protect against accidentally selecting multiple elements.

vec <- c(10, 20, 30, 40, 50)
vec[[2]]    # 20
vec[[2:3]]  # Error: attempt to select more than one element
vec[[6]]    # Error: subscript out of bounds

lst <- list(10, 20, 30, 40, 50)
lst[[2]]    # 20
lst[[2:3]]  # Error: subscript out of bounds
lst[[6]]    # Error: subscript out of bounds

lst[[i]]    # Error if 'i' does not select exactly one

vec <- c(10, 20, 30, 40, 50)
vec[2]      # c(20)
vec[2:3]    # c(20, 30)
vec[6]      # NA

lst <- list(10, 20, 30, 40, 50)
lst[2]      # list(20)
lst[2:3]    # list(20,30)
lst[6]      # NULL

lst[i]      # Sublist? NULL? NA??

These do the same thing:

temp1 <- read_csv("hugeFile1.csv")
temp2 <- read_csv("hugeFile2.csv")
dframe <- inner_join(temp1, temp2, by = "name")
rm(temp1, temp2)

dframe <- local( {
   temp1 <- read_csv("hugeFile1.csv")
   temp2 <- read_csv("hugeFile2.csv")
   inner_join(temp1, temp2, by = "name")
} )

• local() returns value of last expression (here, the inner_join)
• Variables created within local() are discarded.

The base function data.frame has some quirks.

df <- data.frame(aardvark = "fred")
df$a        # What value, if any, is this?
df$b        # What about this?

The base function data.frame has some quirks.

df <- data.frame(aardvark = "fred")
df$a

## [1] fred
## Levels: fred

df$b

## NULL

Oops. Forgot stringsAsFactors=FALSE.

The dplyr function data_frame is friendlier.

library(dplyr)     # or library(tidyverse)

df <- data_frame(aardvark = "fred")
df$aardvark        # "fred" (character, not factor)
df$a               # NULL; Warning: Unknown or uninitialised column

read_csv is a replacement for read.csv.

library(readr)     # or library(tidyverse)

dframe <- read_csv("file.csv")

• Faster than read.csv for big files
• Does not mangle column names (unlike read.csv)
• Does not convert strings to factors
• Nice progress bar for large files

Use str to quickly reveal data structure.

str(airquality)

## 'data.frame':    153 obs. of  6 variables:
##  $ Ozone  : int  41 36 12 18 NA 28 23 19 8 NA ...
##  $ Solar.R: int  190 118 149 313 NA NA 299 99 19 194 ...
##  $ Wind   : num  7.4 8 12.6 11.5 14.3 14.9 8.6 13.8 20.1 8.6 ...
##  $ Temp   : int  67 72 74 62 56 66 65 59 61 69 ...
##  $ Month  : int  5 5 5 5 5 5 5 5 5 5 ...
##  $ Day    : int  1 2 3 4 5 6 7 8 9 10 ...

The zealot package provides an unpacking or destructuring operator, written %<-%, similar to Python or Matlab. These are identical:

temp <- functionThatReturnsAList()
x <- temp[[1]]
y <- temp[[2]]
rm(temp)

library(zealot)
c(x, y) %<-% functionThatReturnsAList()

Has many, many features. Check the help page.

Histograms of small samples aren’t useful. Use stripchart instead to get a sense of the data.

stripchart(samp)

samp <- rgamma(1000, 2, 3)
MASS::fitdistr(samp, "gamma")

##      shape       rate   
##   2.1744783   3.3397465 
##  (0.0907772) (0.1567427)

• Very handy for quick, parametric Monte Carlo or bootstrap, for example
• Understands many distributions: beta, chi-squared, exponential, gamma, geometric, normal, Poisson, t, and many more
• See help page for full list of distributions

Instead of printing the full summary(m) of a regression, you can pull individual regression statistics.

m <- lm(dist ~ speed, data=cars)
summary(m)$adj.r.squared

## [1] 0.6438102

summary(m)$coefficients

##               Estimate Std. Error   t value     Pr(>|t|)
## (Intercept) -17.579095  6.7584402 -2.601058 1.231882e-02
## speed         3.932409  0.4155128  9.463990 1.489836e-12

• residuals
• coefficients
• sigma
• df (degrees of freedom)
• r.squared
• adj.r.squared
• fstatistic (F statistic)
• cov.unscaled (Unscaled variance/covariance)

What gets assigned to x?

x <- dframe %>%
       group_by(country) %>%
       mutate(country_total = sum(population)) %>%
       ungroup %>%
       mutate(country_pct = revenue / country_total)

Ending with an explicit select makes the result obvious.

x <- dframe %>%
       group_by(country) %>%
       mutate(country_total = sum(population)) %>%
       ungroup %>%
       mutate(country_pct = population / country_total) %>%
       select(date, country, population, country_pct)

with(dframe, expr) means use columns of dframe to evaluation expr. For example, these are equivlent.

cars$speed * cars$dist
with(cars, speed * dist)

Handy for pulling out columns at end of pipline.

# Want median diff. between 'after' and 'before' columns
full_join(df1, df2, by="name") %>%
    with(median(after - before))

Suppose you’re working in RStudio . .

• Ctrl + Shift + 1
• Via menu: View -> Panes -> Zoom Source

• Ctrl + Shift + 2
• Via menu: View -> Panes -> Zoom Console

• View(dframe) opens data viewer
• Pop-out the data viewer
• Changes to dframe are immediately seen in pop-out window
• Great for teaching and demonstrations
• Handy for incrementally debugging pipeline transformations
• Sad note: Does not work with data.table

• Start with any vanilla R script
• From RStudio menu, File -> Compile Report …
• Creates a simple HTML or PDF document showing code and output, including graphics
• Easy way to share your results

• Capture your thought process and your data process in a script; e.g.
  • Load and prepare data
  • Estimate model
  • Report results
• The workspace is merely the artifact of running the script.
• The script becomes the final definition of your work.
• So save the script, not the workspace.
• Critical for reproducible results.