diff options
Diffstat (limited to 'r.html.markdown')
| -rw-r--r-- | r.html.markdown | 289 |
1 files changed, 145 insertions, 144 deletions
diff --git a/r.html.markdown b/r.html.markdown index 79af40ce..2746d1eb 100644 --- a/r.html.markdown +++ b/r.html.markdown @@ -4,6 +4,7 @@ contributors: - ["e99n09", "http://github.com/e99n09"] - ["isomorphismes", "http://twitter.com/isomorphisms"] - ["kalinn", "http://github.com/kalinn"] + - ["mribeirodantas", "http://github.com/mribeirodantas"] filename: learnr.r --- @@ -29,13 +30,13 @@ R is a statistical computing language. It has lots of libraries for uploading an # R without understanding anything about programming. Do not worry # about understanding everything the code does. Just enjoy! -data() # browse pre-loaded data sets -data(rivers) # get this one: "Lengths of Major North American Rivers" -ls() # notice that "rivers" now appears in the workspace -head(rivers) # peek at the data set +data() # browse pre-loaded data sets +data(rivers) # get this one: "Lengths of Major North American Rivers" +ls() # notice that "rivers" now appears in the workspace +head(rivers) # peek at the data set # 735 320 325 392 524 450 -length(rivers) # how many rivers were measured? +length(rivers) # how many rivers were measured? # 141 summary(rivers) # what are some summary statistics? # Min. 1st Qu. Median Mean 3rd Qu. Max. @@ -91,14 +92,15 @@ stem(log(rivers)) # Notice that the data are neither normal nor log-normal! # 82 | 2 # make a histogram: -hist(rivers, col="#333333", border="white", breaks=25) # play around with these parameters -hist(log(rivers), col="#333333", border="white", breaks=25) # you'll do more plotting later +hist(rivers, col = "#333333", border = "white", breaks = 25) +hist(log(rivers), col = "#333333", border = "white", breaks = 25) +# play around with these parameters, you'll do more plotting later # Here's another neat data set that comes pre-loaded. R has tons of these. data(discoveries) -plot(discoveries, col="#333333", lwd=3, xlab="Year", +plot(discoveries, col = "#333333", lwd = 3, xlab = "Year", main="Number of important discoveries per year") -plot(discoveries, col="#333333", lwd=3, type = "h", xlab="Year", +plot(discoveries, col = "#333333", lwd = 3, type = "h", xlab = "Year", main="Number of important discoveries per year") # Rather than leaving the default ordering (by year), @@ -109,7 +111,7 @@ sort(discoveries) # [51] 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 # [76] 4 4 4 4 5 5 5 5 5 5 5 6 6 6 6 6 6 7 7 7 7 8 9 10 12 -stem(discoveries, scale=2) +stem(discoveries, scale = 2) # # The decimal point is at the | # @@ -134,7 +136,7 @@ summary(discoveries) # 0.0 2.0 3.0 3.1 4.0 12.0 # Roll a die a few times -round(runif(7, min=.5, max=6.5)) +round(runif(7, min = .5, max = 6.5)) # 1 4 6 1 4 6 4 # Your numbers will differ from mine unless we set the same random.seed(31337) @@ -157,69 +159,68 @@ rnorm(9) # INTEGERS # Long-storage integers are written with L -5L # 5 -class(5L) # "integer" +5L # 5 +class(5L) # "integer" # (Try ?class for more information on the class() function.) # In R, every single value, like 5L, is considered a vector of length 1 -length(5L) # 1 +length(5L) # 1 # You can have an integer vector with length > 1 too: -c(4L, 5L, 8L, 3L) # 4 5 8 3 -length(c(4L, 5L, 8L, 3L)) # 4 -class(c(4L, 5L, 8L, 3L)) # "integer" +c(4L, 5L, 8L, 3L) # 4 5 8 3 +length(c(4L, 5L, 8L, 3L)) # 4 +class(c(4L, 5L, 8L, 3L)) # "integer" # NUMERICS # A "numeric" is a double-precision floating-point number -5 # 5 -class(5) # "numeric" +5 # 5 +class(5) # "numeric" # Again, everything in R is a vector; # you can make a numeric vector with more than one element -c(3,3,3,2,2,1) # 3 3 3 2 2 1 +c(3, 3, 3, 2, 2, 1) # 3 3 3 2 2 1 # You can use scientific notation too -5e4 # 50000 -6.02e23 # Avogadro's number -1.6e-35 # Planck length +5e4 # 50000 +6.02e23 # Avogadro's number +1.6e-35 # Planck length # You can also have infinitely large or small numbers -class(Inf) # "numeric" -class(-Inf) # "numeric" +class(Inf) # "numeric" +class(-Inf) # "numeric" # You might use "Inf", for example, in integrate(dnorm, 3, Inf); # this obviates Z-score tables. # BASIC ARITHMETIC # You can do arithmetic with numbers # Doing arithmetic on a mix of integers and numerics gives you another numeric -10L + 66L # 76 # integer plus integer gives integer -53.2 - 4 # 49.2 # numeric minus numeric gives numeric -2.0 * 2L # 4 # numeric times integer gives numeric -3L / 4 # 0.75 # integer over numeric gives numeric -3 %% 2 # 1 # the remainder of two numerics is another numeric +10L + 66L # 76 # integer plus integer gives integer +53.2 - 4 # 49.2 # numeric minus numeric gives numeric +2.0 * 2L # 4 # numeric times integer gives numeric +3L / 4 # 0.75 # integer over numeric gives numeric +3 %% 2 # 1 # the remainder of two numerics is another numeric # Illegal arithmetic yields you a "not-a-number": -0 / 0 # NaN -class(NaN) # "numeric" +0 / 0 # NaN +class(NaN) # "numeric" # You can do arithmetic on two vectors with length greater than 1, # so long as the larger vector's length is an integer multiple of the smaller -c(1,2,3) + c(1,2,3) # 2 4 6 +c(1, 2, 3) + c(1, 2, 3) # 2 4 6 # Since a single number is a vector of length one, scalars are applied # elementwise to vectors -(4 * c(1,2,3) - 2) / 2 # 1 3 5 +(4 * c(1, 2, 3) - 2) / 2 # 1 3 5 # Except for scalars, use caution when performing arithmetic on vectors with # different lengths. Although it can be done, -c(1,2,3,1,2,3) * c(1,2) # 1 4 3 2 2 6 -# Matching lengths is better practice and easier to read -c(1,2,3,1,2,3) * c(1,2,1,2,1,2) +c(1, 2, 3, 1, 2, 3) * c(1, 2) # 1 4 3 2 2 6 +# Matching lengths is better practice and easier to read most times +c(1, 2, 3, 1, 2, 3) * c(1, 2, 1, 2, 1, 2) # 1 4 3 2 2 6 # CHARACTERS # There's no difference between strings and characters in R -"Horatio" # "Horatio" -class("Horatio") # "character" -class('H') # "character" +"Horatio" # "Horatio" +class("Horatio") # "character" +class("H") # "character" # Those were both character vectors of length 1 # Here is a longer one: -c('alef', 'bet', 'gimmel', 'dalet', 'he') -# => -# "alef" "bet" "gimmel" "dalet" "he" +c("alef", "bet", "gimmel", "dalet", "he") +# => "alef" "bet" "gimmel" "dalet" "he" length(c("Call","me","Ishmael")) # 3 # You can do regex operations on character vectors: -substr("Fortuna multis dat nimis, nulli satis.", 9, 15) # "multis " +substr("Fortuna multis dat nimis, nulli satis.", 9, 15) # "multis " gsub('u', 'ø', "Fortuna multis dat nimis, nulli satis.") # "Fortøna møltis dat nimis, nølli satis." # R has several built-in character vectors: letters @@ -230,32 +231,33 @@ month.abb # "Jan" "Feb" "Mar" "Apr" "May" "Jun" "Jul" "Aug" "Sep" "Oct" "Nov" "D # LOGICALS # In R, a "logical" is a boolean -class(TRUE) # "logical" -class(FALSE) # "logical" + +class(TRUE) # "logical" +class(FALSE) # "logical" # Their behavior is normal -TRUE == TRUE # TRUE -TRUE == FALSE # FALSE -FALSE != FALSE # FALSE -FALSE != TRUE # TRUE +TRUE == TRUE # TRUE +TRUE == FALSE # FALSE +FALSE != FALSE # FALSE +FALSE != TRUE # TRUE # Missing data (NA) is logical, too -class(NA) # "logical" +class(NA) # "logical" # Use | and & for logic operations. # OR -TRUE | FALSE # TRUE +TRUE | FALSE # TRUE # AND -TRUE & FALSE # FALSE +TRUE & FALSE # FALSE # Applying | and & to vectors returns elementwise logic operations -c(TRUE,FALSE,FALSE) | c(FALSE,TRUE,FALSE) # TRUE TRUE FALSE -c(TRUE,FALSE,TRUE) & c(FALSE,TRUE,TRUE) # FALSE FALSE TRUE +c(TRUE, FALSE, FALSE) | c(FALSE, TRUE, FALSE) # TRUE TRUE FALSE +c(TRUE, FALSE, TRUE) & c(FALSE, TRUE, TRUE) # FALSE FALSE TRUE # You can test if x is TRUE -isTRUE(TRUE) # TRUE +isTRUE(TRUE) # TRUE # Here we get a logical vector with many elements: -c('Z', 'o', 'r', 'r', 'o') == "Zorro" # FALSE FALSE FALSE FALSE FALSE -c('Z', 'o', 'r', 'r', 'o') == "Z" # TRUE FALSE FALSE FALSE FALSE +c("Z", "o", "r", "r", "o") == "Zorro" # FALSE FALSE FALSE FALSE FALSE +c("Z", "o", "r", "r", "o") == "Z" # TRUE FALSE FALSE FALSE FALSE # FACTORS # The factor class is for categorical data -# Factors can be ordered (like childrens' grade levels) or unordered (like colors) +# Factors can be ordered (like grade levels) or unordered (like colors) factor(c("blue", "blue", "green", NA, "blue")) # blue blue green <NA> blue # Levels: blue green @@ -263,31 +265,27 @@ factor(c("blue", "blue", "green", NA, "blue")) # Note that missing data does not enter the levels levels(factor(c("green", "green", "blue", NA, "blue"))) # "blue" "green" # If a factor vector has length 1, its levels will have length 1, too -length(factor("green")) # 1 +length(factor("green")) # 1 length(levels(factor("green"))) # 1 # Factors are commonly seen in data frames, a data structure we will cover later -data(infert) # "Infertility after Spontaneous and Induced Abortion" +data(infert) # "Infertility after Spontaneous and Induced Abortion" levels(infert$education) # "0-5yrs" "6-11yrs" "12+ yrs" # NULL # "NULL" is a weird one; use it to "blank out" a vector -class(NULL) # NULL +class(NULL) # NULL parakeet = c("beak", "feathers", "wings", "eyes") -parakeet -# => -# [1] "beak" "feathers" "wings" "eyes" +parakeet # "beak" "feathers" "wings" "eyes" parakeet <- NULL -parakeet -# => -# NULL +parakeet # NULL # TYPE COERCION # Type-coercion is when you force a value to take on a different type -as.character(c(6, 8)) # "6" "8" -as.logical(c(1,0,1,1)) # TRUE FALSE TRUE TRUE +as.character(c(6, 8)) # "6" "8" +as.logical(c(1,0,1,1)) # TRUE FALSE TRUE TRUE # If you put elements of different types into a vector, weird coercions happen: -c(TRUE, 4) # 1 4 -c("dog", TRUE, 4) # "dog" "TRUE" "4" +c(TRUE, 4) # 1 4 +c("dog", TRUE, 4) # "dog" "TRUE" "4" as.numeric("Bilbo") # => # [1] NA @@ -309,14 +307,15 @@ as.numeric("Bilbo") # VARIABLES # Lots of way to assign stuff: -x = 5 # this is possible -y <- "1" # this is preferred -TRUE -> z # this works but is weird +x = 5 # this is possible +y <- "1" # this is preferred traditionally +TRUE -> z # this works but is weird +# Refer to the Internet for the behaviors and preferences about them. # LOOPS # We've got for loops for (i in 1:4) { - print(i) + print(i) } # We've got while loops a <- 10 @@ -341,11 +340,11 @@ if (4 > 3) { # FUNCTIONS # Defined like so: jiggle <- function(x) { - x = x + rnorm(1, sd=.1) #add in a bit of (controlled) noise + x = x + rnorm(1, sd=.1) # add in a bit of (controlled) noise return(x) } # Called like any other R function: -jiggle(5) # 5±ε. After set.seed(2716057), jiggle(5)==5.005043 +jiggle(5) # 5±ε. After set.seed(2716057), jiggle(5)==5.005043 @@ -357,39 +356,39 @@ jiggle(5) # 5±ε. After set.seed(2716057), jiggle(5)==5.005043 # Let's start from the very beginning, and with something you already know: vectors. vec <- c(8, 9, 10, 11) -vec # 8 9 10 11 +vec # 8 9 10 11 # We ask for specific elements by subsetting with square brackets # (Note that R starts counting from 1) -vec[1] # 8 -letters[18] # "r" -LETTERS[13] # "M" -month.name[9] # "September" -c(6, 8, 7, 5, 3, 0, 9)[3] # 7 +vec[1] # 8 +letters[18] # "r" +LETTERS[13] # "M" +month.name[9] # "September" +c(6, 8, 7, 5, 3, 0, 9)[3] # 7 # We can also search for the indices of specific components, -which(vec %% 2 == 0) # 1 3 +which(vec %% 2 == 0) # 1 3 # grab just the first or last few entries in the vector, -head(vec, 1) # 8 -tail(vec, 2) # 10 11 +head(vec, 1) # 8 +tail(vec, 2) # 10 11 # or figure out if a certain value is in the vector -any(vec == 10) # TRUE +any(vec == 10) # TRUE # If an index "goes over" you'll get NA: -vec[6] # NA +vec[6] # NA # You can find the length of your vector with length() -length(vec) # 4 +length(vec) # 4 # You can perform operations on entire vectors or subsets of vectors -vec * 4 # 16 20 24 28 -vec[2:3] * 5 # 25 30 -any(vec[2:3] == 8) # FALSE +vec * 4 # 32 36 40 44 +vec[2:3] * 5 # 45 50 +any(vec[2:3] == 8) # FALSE # and R has many built-in functions to summarize vectors -mean(vec) # 9.5 -var(vec) # 1.666667 -sd(vec) # 1.290994 -max(vec) # 11 -min(vec) # 8 -sum(vec) # 38 +mean(vec) # 9.5 +var(vec) # 1.666667 +sd(vec) # 1.290994 +max(vec) # 11 +min(vec) # 8 +sum(vec) # 38 # Some more nice built-ins: -5:15 # 5 6 7 8 9 10 11 12 13 14 15 -seq(from=0, to=31337, by=1337) +5:15 # 5 6 7 8 9 10 11 12 13 14 15 +seq(from = 0, to = 31337, by = 1337) # => # [1] 0 1337 2674 4011 5348 6685 8022 9359 10696 12033 13370 14707 # [13] 16044 17381 18718 20055 21392 22729 24066 25403 26740 28077 29414 30751 @@ -397,7 +396,7 @@ seq(from=0, to=31337, by=1337) # TWO-DIMENSIONAL (ALL ONE CLASS) # You can make a matrix out of entries all of the same type like so: -mat <- matrix(nrow = 3, ncol = 2, c(1,2,3,4,5,6)) +mat <- matrix(nrow = 3, ncol = 2, c(1, 2, 3, 4, 5, 6)) mat # => # [,1] [,2] @@ -405,13 +404,13 @@ mat # [2,] 2 5 # [3,] 3 6 # Unlike a vector, the class of a matrix is "matrix", no matter what's in it -class(mat) # => "matrix" +class(mat) # "matrix" "array" # Ask for the first row -mat[1,] # 1 4 +mat[1, ] # 1 4 # Perform operation on the first column -3 * mat[,1] # 3 6 9 +3 * mat[, 1] # 3 6 9 # Ask for a specific cell -mat[3,2] # 6 +mat[3, 2] # 6 # Transpose the whole matrix t(mat) @@ -437,14 +436,14 @@ mat2 # [2,] "2" "cat" # [3,] "3" "bird" # [4,] "4" "dog" -class(mat2) # matrix +class(mat2) # matrix # Again, note what happened! # Because matrices must contain entries all of the same class, # everything got converted to the character class -c(class(mat2[,1]), class(mat2[,2])) +c(class(mat2[, 1]), class(mat2[, 2])) # rbind() sticks vectors together row-wise to make a matrix -mat3 <- rbind(c(1,2,4,5), c(6,7,0,4)) +mat3 <- rbind(c(1, 2, 4, 5), c(6, 7, 0, 4)) mat3 # => # [,1] [,2] [,3] [,4] @@ -458,11 +457,11 @@ mat3 # This data structure is so useful for statistical programming, # a version of it was added to Python in the package "pandas". -students <- data.frame(c("Cedric","Fred","George","Cho","Draco","Ginny"), - c(3,2,2,1,0,-1), - c("H", "G", "G", "R", "S", "G")) +students <- data.frame(c("Cedric", "Fred", "George", "Cho", "Draco", "Ginny"), + c( 3, 2, 2, 1, 0, -1), + c( "H", "G", "G", "R", "S", "G")) names(students) <- c("name", "year", "house") # name the columns -class(students) # "data.frame" +class(students) # "data.frame" students # => # name year house @@ -472,21 +471,22 @@ students # 4 Cho 1 R # 5 Draco 0 S # 6 Ginny -1 G -class(students$year) # "numeric" -class(students[,3]) # "factor" +class(students$year) # "numeric" +class(students[,3]) # "factor" # find the dimensions -nrow(students) # 6 -ncol(students) # 3 -dim(students) # 6 3 -# The data.frame() function converts character vectors to factor vectors -# by default; turn this off by setting stringsAsFactors = FALSE when -# you create the data.frame +nrow(students) # 6 +ncol(students) # 3 +dim(students) # 6 3 +# The data.frame() function used to convert character vectors to factor +# vectors by default; This has changed in R 4.0.0. If your R version is +# older, turn this off by setting stringsAsFactors = FALSE when you +# create the data.frame ?data.frame # There are many twisty ways to subset data frames, all subtly unalike -students$year # 3 2 2 1 0 -1 -students[,2] # 3 2 2 1 0 -1 -students[,"year"] # 3 2 2 1 0 -1 +students$year # 3 2 2 1 0 -1 +students[, 2] # 3 2 2 1 0 -1 +students[, "year"] # 3 2 2 1 0 -1 # An augmented version of the data.frame structure is the data.table # If you're working with huge or panel data, or need to merge a few data @@ -503,19 +503,19 @@ students # note the slightly different print-out # 4: Cho 1 R # 5: Draco 0 S # 6: Ginny -1 G -students[name=="Ginny"] # get rows with name == "Ginny" +students[name == "Ginny"] # get rows with name == "Ginny" # => # name year house # 1: Ginny -1 G -students[year==2] # get rows with year == 2 +students[year == 2] # get rows with year == 2 # => # name year house # 1: Fred 2 G # 2: George 2 G # data.table makes merging two data sets easy # let's make another data.table to merge with students -founders <- data.table(house=c("G","H","R","S"), - founder=c("Godric","Helga","Rowena","Salazar")) +founders <- data.table(house = c("G" , "H" , "R" , "S"), + founder = c("Godric", "Helga", "Rowena", "Salazar")) founders # => # house founder @@ -526,8 +526,8 @@ founders setkey(students, house) setkey(founders, house) students <- founders[students] # merge the two data sets by matching "house" -setnames(students, c("house","houseFounderName","studentName","year")) -students[,order(c("name","year","house","houseFounderName")), with=F] +setnames(students, c("house", "houseFounderName", "studentName", "year")) +students[, order(c("name", "year", "house", "houseFounderName")), with = F] # => # studentName year house houseFounderName # 1: Fred 2 G Godric @@ -538,7 +538,7 @@ students[,order(c("name","year","house","houseFounderName")), with=F] # 6: Draco 0 S Salazar # data.table makes summary tables easy -students[,sum(year),by=house] +students[, sum(year), by = house] # => # house V1 # 1: G 3 @@ -571,7 +571,7 @@ students[studentName != "Draco"] # 5: R Cho 1 # Using data.frame: students <- as.data.frame(students) -students[students$house != "G",] +students[students$house != "G", ] # => # house houseFounderName studentName year # 4 H Helga Cedric 3 @@ -583,13 +583,13 @@ students[students$house != "G",] # Arrays creates n-dimensional tables # All elements must be of the same type # You can make a two-dimensional table (sort of like a matrix) -array(c(c(1,2,4,5),c(8,9,3,6)), dim=c(2,4)) +array(c(c(1, 2, 4, 5), c(8, 9, 3, 6)), dim = c(2, 4)) # => # [,1] [,2] [,3] [,4] # [1,] 1 4 8 3 # [2,] 2 5 9 6 # You can use array to make three-dimensional matrices too -array(c(c(c(2,300,4),c(8,9,0)),c(c(5,60,0),c(66,7,847))), dim=c(3,2,2)) +array(c(c(c(2, 300, 4), c(8, 9, 0)), c(c(5, 60, 0), c(66, 7, 847))), dim = c(3, 2, 2)) # => # , , 1 # @@ -609,7 +609,7 @@ array(c(c(c(2,300,4),c(8,9,0)),c(c(5,60,0),c(66,7,847))), dim=c(3,2,2)) # Finally, R has lists (of vectors) list1 <- list(time = 1:40) -list1$price = c(rnorm(40,.5*list1$time,4)) # random +list1$price = c(rnorm(40, .5*list1$time, 4)) # random list1 # You can get items in the list like so list1$time # one way @@ -682,7 +682,7 @@ write.csv(pets, "pets2.csv") # to make a new .csv file ######################### # Linear regression! -linearModel <- lm(price ~ time, data = list1) +linearModel <- lm(price ~ time, data = list1) linearModel # outputs result of regression # => # Call: @@ -719,7 +719,7 @@ summary(linearModel)$coefficients # another way to extract results # Estimate Std. Error t value Pr(>|t|) # (Intercept) 0.1452662 1.50084246 0.09678975 9.234021e-01 # time 0.4943490 0.06379348 7.74920901 2.440008e-09 -summary(linearModel)$coefficients[,4] # the p-values +summary(linearModel)$coefficients[, 4] # the p-values # => # (Intercept) time # 9.234021e-01 2.440008e-09 @@ -728,8 +728,7 @@ summary(linearModel)$coefficients[,4] # the p-values # Logistic regression set.seed(1) list1$success = rbinom(length(list1$time), 1, .5) # random binary -glModel <- glm(success ~ time, data = list1, - family=binomial(link="logit")) +glModel <- glm(success ~ time, data = list1, family=binomial(link="logit")) glModel # outputs result of logistic regression # => # Call: glm(formula = success ~ time, @@ -745,8 +744,10 @@ glModel # outputs result of logistic regression summary(glModel) # more verbose output from the regression # => # Call: -# glm(formula = success ~ time, -# family = binomial(link = "logit"), data = list1) +# glm( +# formula = success ~ time, +# family = binomial(link = "logit"), +# data = list1) # Deviance Residuals: # Min 1Q Median 3Q Max @@ -780,7 +781,7 @@ plot(linearModel) # Histograms! hist(rpois(n = 10000, lambda = 5), col = "thistle") # Barplots! -barplot(c(1,4,5,1,2), names.arg = c("red","blue","purple","green","yellow")) +barplot(c(1, 4, 5, 1, 2), names.arg = c("red", "blue", "purple", "green", "yellow")) # GGPLOT2 # But these are not even the prettiest of R's plots @@ -788,10 +789,10 @@ barplot(c(1,4,5,1,2), names.arg = c("red","blue","purple","green","yellow")) install.packages("ggplot2") require(ggplot2) ?ggplot2 -pp <- ggplot(students, aes(x=house)) +pp <- ggplot(students, aes(x = house)) pp + geom_bar() ll <- as.data.table(list1) -pp <- ggplot(ll, aes(x=time,price)) +pp <- ggplot(ll, aes(x = time, price)) pp + geom_point() # ggplot2 has excellent documentation (available http://docs.ggplot2.org/current/) |