Working in the tidyverse
Writing files to Google Sheets
This requires authorization, hopefully the Oauth token can be auto-refreshed.
# write this to Google Sheets
pacman::p_load(googlesheets4, googledrive)
# if want to write multiple sheets
my_df <- list(df_name1 = df1, df_name2 = df2)
ss4 <- googlesheets4::gs4_create(
"SHEET NAME",
sheets = my_df
)
Removing empty tibble from a list of tibbles
Filter(nrow, my_list)
# or if list of lists
Filter(length, my_list)
Converting named list to dataframe keeping column with the names
This is using the data.table package
# named list
ll <- mtcars %>%
group_by(cyl) %>%
group_map(~ head(.x, 1L)) %>%
set_names(c("Mazda", "Honda", "Suzuki"))
ll
## $Mazda
## # A tibble: 1 × 10
## mpg disp hp drat wt qsec vs am gear carb
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 22.8 108 93 3.85 2.32 18.6 1 1 4 1
##
## $Honda
## # A tibble: 1 × 10
## mpg disp hp drat wt qsec vs am gear carb
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 21 160 110 3.9 2.62 16.5 0 1 4 4
##
## $Suzuki
## # A tibble: 1 × 10
## mpg disp hp drat wt qsec vs am gear carb
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 18.7 360 175 3.15 3.44 17.0 0 0 3 2
data.table::rbindlist(ll, idcol = 'make') %>% tibble()
## # A tibble: 3 × 11
## make mpg disp hp drat wt qsec vs am gear carb
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 Mazda 22.8 108 93 3.85 2.32 18.6 1 1 4 1
## 2 Honda 21 160 110 3.9 2.62 16.5 0 1 4 4
## 3 Suzuki 18.7 360 175 3.15 3.44 17.0 0 0 3 2
Using setNames to change column names
names <- c("name1", "name2", "name3", "name4")
df <- tibble(
a = sample(letters, 4),
b = sample(1:100, 4),
c = sample(1:100, 4),
d = sample(letters, 4)
)
df %>% setNames(names)
## # A tibble: 4 × 4
## name1 name2 name3 name4
## <chr> <int> <int> <chr>
## 1 o 43 89 o
## 2 i 1 54 i
## 3 l 95 36 y
## 4 d 34 6 n
Renaming columns via pipe
The rename_with() function is very helpful…
df <- tibble(
a_1 = sample(letters, 4),
b_2 = sample(1:100, 4),
c_3 = sample(1:100, 4),
d_4 = sample(letters, 4)
)
df |>
rename_with(~ gsub("\\_.*", "", .) , .cols = everything()) |>
rename_with(~ paste0(., "_24"), .cols = everything())
## # A tibble: 4 × 4
## a_24 b_24 c_24 d_24
## <chr> <int> <int> <chr>
## 1 p 46 66 d
## 2 n 1 7 r
## 3 u 10 25 s
## 4 g 60 40 n
Using map to create named lists
fn <- function(x) {
paste0('test-',x)
}
input <- LETTERS[1:3]
input %>% set_names() %>% map(fn)
## $A
## [1] "test-A"
##
## $B
## [1] "test-B"
##
## $C
## [1] "test-C"
Using rename_with to rename specific columns
Occasionally, you’ll only want to rename certain columns and the rename_with function offers this capability.
df <- tibble(
maa_1 = sample(letters, 4),
maa_2 = sample(1:100, 4),
ma_3 = sample(1:100, 4),
ma_4 = sample(letters, 4)
)
df %>%
rename_with(~str_replace_all(., "maa", "ma"), contains("maa"))
## # A tibble: 4 × 4
## ma_1 ma_2 ma_3 ma_4
## <chr> <int> <int> <chr>
## 1 r 93 44 k
## 2 v 28 58 g
## 3 o 80 81 s
## 4 q 42 1 u
piping and dplyr verbs with lists and purrr
Assume a list with data.frames and I want to use a dplyr verb (or apply any function) within it
list_object %>%
map(~ clean_names(.)) %>%
map(~ mutate(., new_var = colA + colB))
Mutate and Summarise multiple columns
This is well documented here, but a few examples are below. When using across() with we need to wrap the variables in quotes if we’re mentioning multiple. And if we’re using multiple functions they need to be put into a list(). To keep tabs on the results, name the function output.
mtcars %>%
group_by(cyl) %>%
summarise(across(c("disp", "drat", "mpg"), list(mean = mean, sd = sd)))
## # A tibble: 3 × 7
## cyl disp_mean disp_sd drat_mean drat_sd mpg_mean mpg_sd
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 4 105. 26.9 4.07 0.365 26.7 4.51
## 2 6 183. 41.6 3.59 0.476 19.7 1.45
## 3 8 353. 67.8 3.23 0.372 15.1 2.56
The helper functions for select() are also useful for selecting variables. Though note that when using helpers such as contains() that you can only include one string. For instance, this will work.
mtcars %>%
summarise(across(contains("ar"), mean))
## gear carb
## 1 3.6875 2.8125
But this won’t
mtcars %>%
summarise(across(contains("ar|mp"), mean))
## data frame with 0 columns and 1 row
If you want to match across multiple strings, the matches() function will do the trick.
mtcars %>%
summarise(across(matches("ar|mp"), mean))
## mpg gear carb
## 1 20.09062 3.6875 2.8125
Multiple left_joins using dplyr
It’s always possible to use multiple left_join functions, but the easiest way to do merge multiple data sets together may be to put everything into a list and then use the Reduce function. I had used map to work over a lot of data, so everything was in a list. I then took the data I wanted to use as my base and concatenated it to the list.
tmp <- c(list(df), original_list)
Then, using dplyr commands was able to join all of the data. In this case, my original list had 30 separate dataframes.
mass_df <- tmp %>% Reduce(function(df1, df2), left_join(df1, df2), .)
The left_join command can be further defined to specify what we’re joining by, or to select only specific columns that will be joined.
# using "matches" to pull out specific variables
mass_df <- tmp %>%
Reduce(function(df1, df2), left_join(df1, select(df2, matches("avar|bvar"))), .)
# specifying what the join is by
mass_df <- tmp %>% Reduce(function(df1, df2), left_join(df1, df2, by = "index"), .)
Calculating quantiles in tidy fashion
pacman::p_load(tidyverse)
# set up quantiles we want
p <- c(.2, .4, .6, .8)
# create list of functions; one for each quantile; and give names to each
p_names <- map_chr(p, ~paste0('p_',.x*100))
p_funs <- map(p, ~partial(quantile, probs = .x, na.rm = TRUE)) %>%
set_names(nm = p_names)
mtcars %>% group_by(cyl) %>%
summarise(across(mpg, tibble::lst(!!!p_funs))) %>%
janitor::clean_names()
## # A tibble: 3 × 5
## cyl mpg_p_20 mpg_p_40 mpg_p_60 mpg_p_80
## <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 4 22.8 24.4 27.3 30.4
## 2 6 18.3 19.4 20.5 21
## 3 8 13.9 15.0 15.4 16.8
Selection of multiple variables
When using either select() or wanting to mutate(across()) there are lots of helpers – starts_with(), ends_with(), contains(), and matches() and probably some that I’m missing.
contains() only works on a single, specific request, whereas matches() allows for an OR.
tmp <- tibble(g1_letters = sample(letters, 5),
g1_num = sample(1:600000, 5),
g2_letters = sample(letters, 5),
g2_num = sample(1:60000, 5),
h1_letters = sample(letters, 5),
h1_num = sample(1:60000, 5))
# or statement
tmp <- tmp %>% select(matches("g1|h1"))
tmp
## # A tibble: 5 × 4
## g1_letters g1_num h1_letters h1_num
## <chr> <int> <chr> <int>
## 1 h 91961 l 5263
## 2 f 95551 q 58515
## 3 n 203957 a 2203
## 4 t 118099 r 46583
## 5 j 54905 k 39347
But we can also use the intersect() function to create an AND statement
tmp <- tibble(g1_letters = sample(letters, 5),
g1_num = sample(1:600000, 5),
g1_weighted = g1_num*.4,
g2_letters = sample(letters, 5),
g2_num = sample(1:60000, 5),
g2_weighted = g2_num*.4,
h1_letters = sample(letters, 5),
h1_num = sample(1:60000, 5),
h1_weighted = h1_num*.4)
tmp %>%
summarise(across(matches("num|weighted"), sum)) %>%
mutate(across(intersect(starts_with("g1"), contains("weighted")),
~ paste0(., "---->")))
## # A tibble: 1 × 6
## g1_num g1_weighted g2_num g2_weighted h1_num h1_weighted
## <int> <chr> <int> <dbl> <int> <dbl>
## 1 1278472 511388.8----> 185276 74110. 150962 60385.
Piping into a t.test
tibble(a = c(rnorm(100, mean = 50, sd = 5),rnorm(100, mean = 60, sd = 5)),
group = c(rep("green", 100), rep("blue", 100))) %>%
t.test(a ~ group, data = ., var.equal = TRUE)
##
## Two Sample t-test
##
## data: a by group
## t = 11.788, df = 198, p-value < 2.2e-16
## alternative hypothesis: true difference in means between group blue and group green is not equal to 0
## 95 percent confidence interval:
## 7.394397 10.365370
## sample estimates:
## mean in group blue mean in group green
## 59.16839 50.28851
Piping into a cor.test
Method 1 (no group_by)
Using the exposition pipe from the magrittr package.
library(magrittr)
mtcars %$%
cor.test(mpg, hp) %>%
broom::tidy()
## # A tibble: 1 × 8
## estimate statistic p.value parameter conf.low conf.high method alternative
## <dbl> <dbl> <dbl> <int> <dbl> <dbl> <chr> <chr>
## 1 -0.776 -6.74 0.000000179 30 -0.885 -0.586 Pears… two.sided
Method 2 (allows group_by)
Using a nest-map-unnest workflow:
mtcars %>%
nest(data = -cyl) %>%
mutate(test = map(data, ~ cor.test(.x$mpg, .x$hp)), # S3 list-col
tidied = map(test, broom::tidy)
) %>%
unnest(tidied)
## # A tibble: 3 × 11
## cyl data test estimate statistic p.value parameter conf.low conf.high
## <dbl> <list> <list> <dbl> <dbl> <dbl> <int> <dbl> <dbl>
## 1 6 <tibble> <htest> -0.127 -0.286 0.786 5 -0.803 0.692
## 2 4 <tibble> <htest> -0.524 -1.84 0.0984 9 -0.855 0.111
## 3 8 <tibble> <htest> -0.284 -1.02 0.326 12 -0.708 0.291
## # ℹ 2 more variables: method <chr>, alternative <chr>
Arranging within a group
library(tidyverse)
ToothGrowth %>%
group_by(supp) %>%
arrange(len, .by_group = TRUE)
## # A tibble: 60 × 3
## # Groups: supp [2]
## len supp dose
## <dbl> <fct> <dbl>
## 1 8.2 OJ 0.5
## 2 9.4 OJ 0.5
## 3 9.7 OJ 0.5
## 4 9.7 OJ 0.5
## 5 10 OJ 0.5
## 6 14.5 OJ 0.5
## 7 14.5 OJ 1
## 8 15.2 OJ 0.5
## 9 16.5 OJ 0.5
## 10 17.6 OJ 0.5
## # ℹ 50 more rows
Using cross and map to paste
The cross() function is similar to expand.grid. Here, we create a list and use map to paste it together where each in a separate list.
data <- list(qq = "Q",
q = 1:4,
hyphen = "-",
yr = 13:19)
data %>%
cross() %>%
map(lift(paste0)) %>%
head()
## [[1]]
## [1] "Q1-13"
##
## [[2]]
## [1] "Q2-13"
##
## [[3]]
## [1] "Q3-13"
##
## [[4]]
## [1] "Q4-13"
##
## [[5]]
## [1] "Q1-14"
##
## [[6]]
## [1] "Q2-14"
But we can use setNames and then reduce put it into a data.frame.
data <- list(qq = "Q",
q = 1:4,
hyphen = "-",
yr = 13:19)
data %>%
cross() %>%
map(lift(paste0)) %>%
map(setNames, c("QTR")) %>%
reduce(bind_rows) %>%
head()
## # A tibble: 6 × 1
## QTR
## <chr>
## 1 Q1-13
## 2 Q2-13
## 3 Q3-13
## 4 Q4-13
## 5 Q1-14
## 6 Q2-14
Group indices
Assume we have a tibble like so
tibble(letter = rep(letters[1:2], each = 6),
state = c(rep(c(state.abb[c(1,4,5)]), each = 2),
rep(c(state.abb[c(25,38,43)]), each = 2)))
## # A tibble: 12 × 2
## letter state
## <chr> <chr>
## 1 a AL
## 2 a AL
## 3 a AR
## 4 a AR
## 5 a CA
## 6 a CA
## 7 b MO
## 8 b MO
## 9 b PA
## 10 b PA
## 11 b TX
## 12 b TX
and we want to create a group index for each letter. A simple way of doing this is to use factor conversion.
tibble(letter = rep(letters[1:2], each = 6),
state = c(rep(c(state.abb[c(1,4,5)]), each = 2),
rep(c(state.abb[c(25,38,43)]), each = 2))) |>
mutate(id = as.integer(as.factor(letter)))
## # A tibble: 12 × 3
## letter state id
## <chr> <chr> <int>
## 1 a AL 1
## 2 a AL 1
## 3 a AR 1
## 4 a AR 1
## 5 a CA 1
## 6 a CA 1
## 7 b MO 2
## 8 b MO 2
## 9 b PA 2
## 10 b PA 2
## 11 b TX 2
## 12 b TX 2
Group indices within nested groups
Assume we have a data set with a nested group structures like so.
tibble(letter = rep(letters[1:2], each = 6),
state = c(rep(c(state.abb[c(1,4,5)]), each = 2),
rep(c(state.abb[c(25,38,43)]), each = 2)))
## # A tibble: 12 × 2
## letter state
## <chr> <chr>
## 1 a AL
## 2 a AL
## 3 a AR
## 4 a AR
## 5 a CA
## 6 a CA
## 7 b MO
## 8 b MO
## 9 b PA
## 10 b PA
## 11 b TX
## 12 b TX
Our goal is to produce a repeating id for each state within each letter group, so that AL, AR, and CA would be 1, 2, 3 and MO, PA, and TX would also be 1, 2, 3.
But using the group_indices() function doesn’t help us here (note I’m suppressing warnings because I have no idea how to use group_by() first with this function).
tibble(letter = rep(letters[1:2], each = 6),
state = c(rep(c(state.abb[c(1,4,5)]), each = 2),
rep(c(state.abb[c(25,38,43)]), each = 2))) %>%
mutate(id1 = suppressWarnings(group_indices(., letter)),
id2 = suppressWarnings(group_indices(., state)))
## # A tibble: 12 × 4
## letter state id1 id2
## <chr> <chr> <int> <int>
## 1 a AL 1 1
## 2 a AL 1 1
## 3 a AR 1 2
## 4 a AR 1 2
## 5 a CA 1 3
## 6 a CA 1 3
## 7 b MO 2 4
## 8 b MO 2 4
## 9 b PA 2 5
## 10 b PA 2 5
## 11 b TX 2 6
## 12 b TX 2 6
But we can get to what we want by using cumsum and !duplicated…
tibble(letter = rep(letters[1:2], each = 6),
state = c(rep(c(state.abb[c(1,4,5)]), each = 2),
rep(c(state.abb[c(25,38,43)]), each = 2))) %>%
group_by(letter) %>%
mutate(id = cumsum(!duplicated(state)))
## # A tibble: 12 × 3
## # Groups: letter [2]
## letter state id
## <chr> <chr> <int>
## 1 a AL 1
## 2 a AL 1
## 3 a AR 2
## 4 a AR 2
## 5 a CA 3
## 6 a CA 3
## 7 b MO 1
## 8 b MO 1
## 9 b PA 2
## 10 b PA 2
## 11 b TX 3
## 12 b TX 3
Normalize function
This has nothing to do with the tidyverse, but just want to put it here
norm <- function(x) (x - min(x)) / (max(x) - min(x))
