Share
Embed
Expressing yourself with R
Hadley Wickham
September 24, 2017
Transcript
74 slides
-
Tidy Import Visualise Transform Model Program tibble tidyr purrr magrittr
dplyr forcats hms ggplot2 broom modelr readr readxl haven xml2 lubridate stringr tidyverse.org r4ds.had.co.nz -
@aaronwolen, @aghaynes, @ajdamico, @ajschumacher, @alberthkcheng, @alyst, @andrew, @andrewjlm, @apjanke, @arneschillert,
@artemklevtsov, @arunsrinivasan, @asnr, @astamm, @austenhead, @baptiste, @bbolker, @bearloga, @benmarwick, @bhive01, @BioStatMatt, @bpbond, @bquast, @BrianDiggs, @briatte, @burchill, @casallas, @cb4ds, @cboettig, @cderv, @christophergandrud, @cmartin, @colinbrislawn, @coolbutuseless, @cosinequanon, @craigcitro, @csgillespie, @ctbrown, @daattali, @dandermotj, @danliIDEA, @DanRuderman, @davharris, @davidmorrison, @dchiu911, @dchudz, @dewittpe, @dgromer, @dgrtwo, @dhimmel, @dickoa, @diogocp, @djmurphy420, @dlebauer, @dmedri, @dmenne, @dougmitarotonda, @dpastoor, @dpocock, @dtelad11, @earino, @echasnovski, @ecortens, @eddelbuettel, @edgararuiz, @edwindj, @egnha, @ehrlinger, @eibanez, @eipi10, @ekstroem, @emojiencoding, @etiennebr, @evanmiller, @fpinter, @FvD, @gaborcsardi, @gagolews, @garrettgman, @gavinsimpson, @gergness, @gnustats, @gorcha, @goyalmunish, @gregmacfarlane, @guillett, @gvelasq2, @hannesmuehleisen, @has2k1, @helix123, @hmalmedal, @hoehleatsu, @hoesler, @holstius, @hrbrmstr, @ianmcook, @ijlyttle, @ilarischeinin, @imanuelcostigan, @Ironholds, @ismayc, @isomorphisms, @itsdalmo, @JakeRuss, @janschulz, @jasonelaw, @javierluraschi, @jayhesselberth, @jcheng5, @jdnewmil, @jefferis, @jennybc, @jenzopr, @jeremystan, @jeroen, @jgabry, @jhuovari, @jiho, @jimhester, @jirkalewandowski, @jjallaire, @jmarshallnz, @jmi5, @joethorley, @JoFrhwld, @jonboiser, @jonmcalder, @joranE, @joshkatz, @jrnold, @juba, @junkka, @justmarkham, @kalibera, @karawoo, @karthik, @Katiedaisey, @kbenoit, @Kevin-M-Smith, @kevinushey, @kmillar, @kohske, @krlmlr, @kwenzig, @kwstat, @KZARCA, @l-d-s, @LaDilettante, @larmarange, @leondutoit, @lepennec, @lindbrook, @lionel-, @lmullen, @lorenzwalthert, @lselzer, @luckyrandom, @LucyMcGowan, @lwjohnst86, @MarcusWalz, @markdly, @markriseley, @matthieugomez, @maurolepore, @mdlincoln, @mgacc0, @mgirlich, @michaelquinn32, @mikelove, @mkcor, @mkuehn10, @mkuhn, @mmparker, @msonnabaum, @ncarchedi, @NoahMarconi, @noamross, @npjc, @nutterb, @paternogbc, @paul-buerkner, @PedramNavid, @PeteHaitch, @pierucci, @pimentel, @pitakakariki, @pkq, @r2evans, @rbdixon, @richierocks, @RiRam, @rmsharp, @robertzk, @rohan-shah, @romainfrancois, @RoyalTS, @rsaporta, @rtaph, @rudazhan, @ruderphilipp, @s-fleck, @seaaan, @setempler, @sfirke, @shabbybanks, @sjackman, @sjPlot, @smbache, @statisfactions, @steromano, @t-kalinowski, @tareefk, @tdhock, @terrytangyuan, @thomasp85, @tjmahr, @tklebel, @tmshn, @tonytonov, @tuttinator, @tverbeke, @uribo, @vspinu, @wch, @webbedfeet, @wibeasley, @wligtenberg, @x0rshift, @xiaodaigh, @Yeedle, @yutannihilation, @zeehio, @zhaoy, and @zhilongjia -
Solve complex problems by combining simple pieces that have a
consistent structure -
df$a[df$a == -99] <- NA df$b[df$b == -99] <- NA
df$e[df$e == -99] <- NA df$f[df$f == -99] <- NA df$g[df$g == -98] <- NA df$h[df$h == -99] <- NA df$i[df$i == -99] <- NA df$i[df$j == -99] <- NA df$k[df$k == -99] <- NA df$l[df$l == -99] <- NA df$m[df$m == -99] <- NA df$n[df$n == -99] <- NA What’s the point of this code? What’s wrong? -
df$a[df$a == -99] <- NA df$b[df$b == -99] <- NA
# c & d are character variables df$e[df$e == -99] <- NA df$f[df$f == -99] <- NA df$g[df$g == -98] <- NA df$h[df$h == -99] <- NA df$i[df$i == -99] <- NA df$i[df$j == -99] <- NA df$k[df$k == -99] <- NA df$l[df$l == -99] <- NA df$m[df$m == -99] <- NA df$n[df$n == -99] <- NA Duplicated code hides intent & errors -
fix_missing <- function(x) { x[x == -99] <- NA x
} df$a <- fix_missing(df$a) df$b <- fix_missing(df$b) df$e <- fix_missing(df$e) df$f <- fix_missing(df$f) df$g <- fix_missing(df$g) df$h <- fix_missing(df$h) df$i <- fix_missing(df$i) df$j <- fix_missing(df$j) df$k <- fix_missing(df$k) Create a function whenever you’ve pasted >3 times -
fix_missing <- function(x) { x[x == -99] <- NA x
} df <- purrr::modify_if(df, is.numeric, fix_missing) Learn FP tools to remove even more duplication -
What is a simple function? Does one thing well Needs
minimal context to be understood -
Computes a value / Changes the world One thing well
Minimal context Type stable Obey scoping rules No hidden arguments Evocative name -
Computes a value / Changes the world One thing well
Minimal context Type stable Obey scoping rules No hidden arguments Evocative name -
# Computes a value mean() mutate() + geom_line() # Changes
the world print() write_csv() <- Which is which? -
runif(5) #> [1] 0.5530 0.0138 0.8774 0.9225 0.0606 runif(5) #>
[1] 0.8210 0.0459 0.6008 0.4323 0.3644 Some functions must do both -
.Random.seed[2] #> [1] 624 runif(5) #> [1] 0.0808 0.8343 0.6008
0.1572 0.0074 .Random.seed[2] #> [1] 5 Some functions must do both -
mod <- lm(mpg ~ wt, data = mtcars) summary(mod) #>
Coefficients: #> Estimate Std. Error t value Pr(>|t|) #> (Intercept) 37.285 1.878 19.86 < 2e-16 *** #> wt -5.344 0.559 -9.56 1.3e-10 *** #> --- #> #> Residual standard error: 3.05 on 30 degrees of freedom #> Multiple R-squared: 0.753, Adjusted R-squared: 0.745 #> F-statistic: 91.4 on 1 and 30 DF, p-value: 1.29e-10 Base R generally does this well -
So the exceptions are extra frustrating 10 15 20 25
30 −4 0 2 4 6 8 Fitted values Residuals • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • Residuals vs Fitted Fiat 128 Toyota Corolla Chrysler Imperial • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • −2 −1 0 1 2 −1 0 1 2 Theoretical Quantiles Standardized residuals Normal Q−Q Fiat 128 Toyota Corolla Chrysler Imperial 10 15 20 25 30 0.0 0.5 1.0 1.5 Fitted values Standardized residuals • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • Scale−Location Fiat 128 Toyota Corolla Chrysler Imperial 0.00 0.05 0.10 0.15 0.20 −2 −1 0 1 2 Leverage Standardized residuals • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • Cook's distance 0.5 0.5 1 Residuals vs Leverage Chrysler Imperial Toyota Corolla Fiat 128 -
“Mama always said type- unstable functions are like a box
of chocolates. You never know what you’re gonna get.” — Hadley Gump -
Type-stable functions f f f g g g Regardless of
the input, a type-stable function gives the same type of output It’s harder to predict the result of a type-unstable function -
# A tibble: 150 x 5 Sepal.Length Sepal.Width Petal.Length Petal.Width
Species <dbl> <dbl> <dbl> <dbl> <fctr> 1 5.1 3.5 1.4 0.2 setosa 2 4.9 3.0 1.4 0.2 setosa 3 4.7 3.2 1.3 0.2 setosa 4 4.6 3.1 1.5 0.2 setosa 5 5.0 3.6 1.4 0.2 setosa 6 5.4 3.9 1.7 0.4 setosa 7 4.6 3.4 1.4 0.3 setosa 8 5.0 3.4 1.5 0.2 setosa 9 4.4 2.9 1.4 0.2 setosa 10 4.9 3.1 1.5 0.1 setosa # ... with 140 more rows iris -
find_vars <- function(df, predicate) { vars <- sapply(df, predicate) df[,
vars] } find_vars(iris, is.numeric) find_vars(iris, is.factor) # For experts only: find_vars(iris[, 0], is.numeric) What will this function return? iris has four numeric variables and one factor -
class(find_vars(iris, is.numeric)) #> [1] "data.frame" class(find_vars(iris, is.factor)) #> [1] "factor"
find_vars(iris[, 0], is.numeric) #> Error in .subset(x, j): #> invalid subscript type 'list' -
find_vars <- function(df, predicate) { vars <- sapply(df, predicate) df[,
vars] } sapply() & [.data.frame are type-unstable Returns vector or data frame Returns vector, matrix, or list -
find_vars <- function(df, predicate) { vars <- purrr::map_lgl(df, predicate) df[,
vars, drop = FALSE] } Two changes make it much more predictable -
by_dest <- group_by(flights, dest) dest_delay <- summarise(by_dest, delay = mean(dep_delay,
na.rm = TRUE), n = n() ) big_dest <- filter(dest_delay, n > 100) arrange(big_dest, desc(delay)) Base R has two ways to combine functions -
foo <- group_by(flights, dest) foo <- summarise(foo, delay = mean(dep_delay,
na.rm = TRUE), n = n() ) foo <- filter(foo, n > 100) arrange(foo, desc(delay)) But naming is hard work -
foo1 <- group_by(flights, dest) foo2 <- summarise(foo1, delay = mean(dep_delay,
na.rm = TRUE), n = n() ) foo3 <- filter(foo2, n > 100) arrange(foo2, desc(delay)) But naming is hard work -
arrange( filter( summarise( group_by(flights, dest), delay = mean(dep_delay, na.rm =
TRUE), n = n() ), n > 100 ), desc(delay) ) Alternatively, you could nest function calls -
x %>% f() # Is the same as f(x) x
%>% f() %>% g(y) # Is the same as g(f(x), y) The pipe -
flights %>% group_by(dest) %>% summarise( delay = mean(dep_delay, na.rm =
TRUE), n = n() ) %>% filter(n > 100) %>% arrange(desc(delay)) This is easy to read & doesn’t require naming -
library(tidyverse) library(magick) dir(pattern = ".png") %>% map(image_read) %>% image_join() %>%
image_animate(fps = 1, loop = 25) %>% image_write("my_animation.gif") Makes it easy to read unfamiliar code https://twitter.com/ricardokriebel/status/849626401611411458 What does this code do? -
Read left-to-right Omits intermediate values Non-linear y <- f(x)
g(y) ✅ ✅ g(f(x)) ✅ ✅ x %>% f() %>% g() ✅ ✅ -
flights %>% group_by(date) %>% summarise(n = n()) %>% ggplot(aes(date, n))
+ geom_line() What happens if your pieces aren’t simple functions? -
ggsave( flights %>% group_by(date) %>% summarise(n = n()) %>% ggplot(aes(date,
n)) + geom_line(), "my-plot.pdf" ) Which makes it quite inconsistent -
# https://github.com/hadley/ggplot1 library(ggplot1) flights %>% group_by(date) %>% summarise(n = n())
%>% ggplot(aes(date, n)) %>% ggpoint() %>% ggsave("my-plot.pdf") Interestingly, ggplot did not have this problem -
flights %>% group_by(dest) %>% summarise( delay = mean(dep_delay, na.rm =
TRUE), n = n() ) %>% filter(n > 100) %>% arrange(desc(delay)) -> dest_delays Another interesting connection is -> -
dest_delays <- flights %>% group_by(dest) %>% summarise( delay = mean(dep_delay,
na.rm = TRUE), n = n() ) %>% filter(n > 100) %>% arrange(desc(delay)) But leading with assignment improves readability -
Tidy data is a consistent way of storing data 1.
Each dataset goes in a data frame. 2. Each variable goes in a column. -
Tidy datasets are all alike; every messy dataset is
messy in its own way — Hadley Tolstoy -
# A tibble: 5,769 × 22 iso2 year m04 m514
m014 m1524 m2534 m3544 m4554 m5564 m65 mu f04 f514 f014 f1524 <chr> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> 1 AD 1989 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 2 AD 1990 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 3 AD 1991 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 4 AD 1992 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 5 AD 1993 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 6 AD 1994 NA NA NA NA NA NA NA NA NA NA NA NA NA NA 7 AD 1996 NA NA 0 0 0 4 1 0 0 NA NA NA 0 1 8 AD 1997 NA NA 0 0 1 2 2 1 6 NA NA NA 0 1 9 AD 1998 NA NA 0 0 0 1 0 0 0 NA NA NA NA NA 10 AD 1999 NA NA 0 0 0 1 1 0 0 NA NA NA 0 0 11 AD 2000 NA NA 0 0 1 0 0 0 0 NA NA NA NA NA 12 AD 2001 NA NA 0 NA NA 2 1 NA NA NA NA NA NA NA 13 AD 2002 NA NA 0 0 0 1 0 0 0 NA NA NA 0 1 14 AD 2003 NA NA 0 0 0 1 2 0 0 NA NA NA 0 1 15 AD 2004 NA NA 0 0 0 1 1 0 0 NA NA NA 0 0 16 AD 2005 0 0 0 0 1 1 0 0 0 0 0 0 0 1 17 AD 2006 0 0 0 1 1 2 0 1 1 0 0 0 0 0 # ... with 5,752 more rows, and 6 more variables: f2534 <int>, f3544 <int>, f4554 <int>, # f5564 <int>, f65 <int>, fu <int> Messy data has a varied shape What are the variables in this dataset? (Hint: f = female, u = unknown, 1524 = 15-24) -
# A tibble: 35,750 × 5 country year sex age
n <chr> <int> <chr> <chr> <int> 1 AD 1996 f 014 0 2 AD 1996 f 1524 1 3 AD 1996 f 2534 1 4 AD 1996 f 3544 0 5 AD 1996 f 4554 0 6 AD 1996 f 5564 1 7 AD 1996 f 65 0 8 AD 1996 m 014 0 9 AD 1996 m 1524 0 10 AD 1996 m 2534 0 # ... with 35,740 more rows Tidy data has a uniform shape -
The family of Dashwood had long been settled in Sussex.
Their estate was large, and their residence was at Norland Park, in the centre of their property, where, for many generations, they had lived in so respectable a manner as to engage the general good opinion of their surrounding acquaintance. — Sense & Sensibility, Jane Austen -
# A tibble: 724,880 × 4 book linenumber chapter word
<fctr> <int> <int> <chr> 1 Sense & Sensibility 10 1 chapter 2 Sense & Sensibility 10 1 1 3 Sense & Sensibility 13 1 the 4 Sense & Sensibility 13 1 family 5 Sense & Sensibility 13 1 of 6 Sense & Sensibility 13 1 dashwood 7 Sense & Sensibility 13 1 had 8 Sense & Sensibility 13 1 long 9 Sense & Sensibility 13 1 been 10 Sense & Sensibility 13 1 settled # ... with 724,870 more rows tidytext provides an answer -
Emma Northanger Abbey Persuasion Sense & Sensibility Pride & Prejudice
Mansfield Park 0 50 100 150 0 20 40 60 80 0 20 40 60 80 0 50 100 0 50 100 0 50 100 150 −50 −25 0 25 50 −50 −25 0 25 50 sentiment Sentiment of Jane Austen books -
34°N 34.5°N 35°N 35.5°N 36°N 36.5°N 84°W 82°W 80°W 78°W
76°W 84°W 82°W 80°W 78°W 76°W 34°N 34.5°N 35°N 35.5°N 36°N 36.5°N 0.05 0.10 0.15 0.20 AREA -
nc <- sf::st_read(system.file("shape/nc.shp", package = "sf")) nc %>% as_tibble() %>%
select(NAME, FIPS, AREA, geometry) #> # A tibble: 100 × 4 #> NAME FIPS AREA geometry #> <fctr> <fctr> <dbl> <simple_feature> #> 1 Ashe 37009 0.114 <MULTIPOLYGON...> #> 2 Alleghany 37005 0.061 <MULTIPOLYGON...> #> 3 Surry 37171 0.143 <MULTIPOLYGON...> #> 4 Currituck 37053 0.070 <MULTIPOLYGON...> #> 5 Northampton 37131 0.153 <MULTIPOLYGON...> #> 6 Hertford 37091 0.097 <MULTIPOLYGON...> #> 7 Camden 37029 0.062 <MULTIPOLYGON...> #> 8 Gates 37073 0.091 <MULTIPOLYGON...> #> 9 Warren 37185 0.118 <MULTIPOLYGON...> #> 10 Stokes 37169 0.124 <MULTIPOLYGON...> #> # ... with 90 more rows Store complex geometries in a list-column -
What if you have complex data? 1. Each dataset goes
in a tibble. 2. Each variable goes in a column. -
df <- tibble(xyz = "a") df$x #> Warning: Unknown column
'x' #> NULL df$xyz #> [1] "a" Tibbles are data frames that are lazy & surly -
data.frame(x = list(1:2, 3:5)) #> Error: arguments imply differing number
#> of rows: 2, 3 tibble(x = list(1:2, 3:5)) #> # A tibble: 2 x 1 #> x #> <list> #> 1 <int [2]> #> 2 <int [3]> But also have better support for list-cols -
List-columns keep related things together Anything can go in a
list & a list can go in a data frame -
Solve complex problems by combining simple pieces that have a
consistent structure -
Solve complex problems by combining simple pieces that have a
consistent structure Functions that do one thing well & can be understood with minimal context -
Solve complex problems by combining simple pieces that have a
consistent structure With assignment, composition, or the pipe -
Solve complex problems by combining simple pieces that have a
consistent structure Tidy tibbles have variables in columns and cases in rows. List-cols can store richer data structures -
Tidy Import Visualise Transform Model Program tibble tidyr purrr magrittr
dplyr forcats hms ggplot2 broom modelr readr readxl haven xml2 lubridate stringr tidyverse.org r4ds.had.co.nz -
This work is licensed under the Creative Commons Attribution-Noncommercial
3.0 United States License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/3.0/us/