Open Online R Stream - 2020/12/10
News, Resources, Errata/Additions

---

R Related News

• Native pipe operator coming to R
  https://stat.ethz.ch/R-manual/R-devel/doc/html/NEWS.html
  

  Note: The following code will only work in the development version of R (R-devel)

  sqrt(log(5))
  
  # install the tidyverse packages
  #install.packages(tidyverse)
  
  # using the 'magrittr' pipe
  library(magrittr)
  5 %>% log() %>% sqrt()
  
  # using the native R pipe
  5 |> log() |> sqrt()
  
  # note: this works
  5 %>% log %>% sqrt
  
  # this does not
  5 |> log |> sqrt

• There will also be a new shorthand notation for writing functions
  \(x) x + 1 is then a shorter way of writing function(x) x + 1
  This is useful when writing ‘anonymous’ functions

  # create some toy data
  set.seed(123)
  dat <- data.frame(id = rep(1:4, each=5),
                    y  = sample(5:10, 4*5, replace=TRUE))
  dat
  
  # what is the most frequent response (i.e., mode) for each subject?
  # note: R does not have a built-in function to obtain the mode!
  # the mode() function does not compute the mode, but gets the object type
  
  mode(dat$y)
  
  # we could do the following
  
  table(dat$y)
  which.max(table(dat$y))
  names(which.max(table(dat$y)))
  as.numeric(names(which.max(table(dat$y))))
  
  # note: there is a better way of doing this that works for both numeric and
  # character/factor variables (https://stackoverflow.com/a/8189441/2615367);
  
  # we now want the mode for every subject; we can use by() for this
  
  # for example, to get the mean of every subject, we would do
  by(dat$y, dat$id, mean)
  
  # but there is no function for the mode, so we could write this outselves
  getmode <- function(x)
     as.numeric(names(which.max(table(x))))
  
  # then we can do
  by(dat$y, dat$id, getmode)
  
  # but we may not need getmode() any further, so it clutters up our workspace;
  
  # we can remove it
  rm(getmode)
  
  # an alternative approach is to write an anonymous function (i.e., we just
  # create the function we need 'on the fly' within our by() call)
  by(dat$y, dat$id, function(x) as.numeric(names(which.max(table(x)))))
  
  # and to save a little bit of typing, we can use the `\(x)` syntax
  by(dat$y, dat$id, \(x) as.numeric(names(which.max(table(x)))))

  As it says in the NEWS file: “These features are experimental and may change prior to release.” Also, there is some active discussion going on (around the native pipe and this new syntax for creating functions) on the R-devel mailing list (https://stat.ethz.ch/mailman/listinfo/r-devel).

---

R Related Resources

---

Errata / Additions

• last time we looked at some packages from the tidyverse (https://www.tidyverse.org)

• as noted above, magrittr provides the tidyverse pipe

• this can be combined with dplyr for various data preparation/manipulation steps

  library(tidyverse)
  
  # create some toy data
  set.seed(123)
  dat <- data.frame(id = rep(1:4, each=5),
                    y  = sample(5:10, 4*5, replace=TRUE))
  dat
  
  # get the mean of every subject the tidyverse way
  dat %>%
     group_by(id) %>%
     summarise(meany = mean(y))

• a lightweight alternative to dplyr and magrittr is provided by the poorman package (https://cran.r-project.org/package=poorman)

  # create some toy data
  set.seed(123)
  dat <- data.frame(id = rep(1:4, each=5),
                    y  = sample(5:10, 4*5, replace=TRUE))
  dat
  
  # install and load the poorman package
  #install.packages(poorman)
  library(poorman)
  
  # get the mean of every subject
  dat %>%
     group_by(id) %>%
     summarise(meany = mean(y))

• and in the future, should be able to combine this with R’s native pipe

  dat |>
     group_by(id) |>
     summarise(meany = mean(y))

---