lab 3 · functions (R) — source
A. Stamm 
This page shows the source file of the lab below. Use the button to download it as a plain file and run it in your own workspace.
# ============================================================
# Lab 3: Writing custom functions
# The R Statistical Language · Session 3
#
# Work through the exercises. Comment your thought process
# as you go: a reproducible script is one a stranger can read.
# ============================================================
## Part A: first functions ---------------------------------
# A1. Write zscore <- function(x) returning (x - mean(x)) / sd(x).
# Test it on c(1, 2, 3, 4, 5): one value should be negative,
# one positive, and the mean of the output 0.
# A2. Write mad_ratio(x) = mean(abs(x - mean(x))) / sd(x).
# Test it on rnorm(1000). The result should be close to 0.8.
# (mad_ratio is a robust/modern measure relative to the sd.)
# A3. Write describe(x) that returns a named list with mean, sd,
# median, min, max. Test it on a rnorm(100) sample.
## Part B: arguments & defaults ---------------------------------
# B1. Write clamp(x, lo = 0, hi = 1) using ifelse(). Verify that
# clamp(c(-1, 0.5, 2)) == c(0, 0.5, 1).
# B2. Write a function temp_label(celsius) with an argument
# scale = "celsius" that returns a label vector with ifelse()
# thresholds. Try switching the scale to interpret Fahrenheit.
## Part C: functions + pipelines ---------------------------------
# C1. Load the penguins data. Write a function avg_mass_by(df, group)
# that groups by a bare column name and returns the mean body mass.
# (Hint: use filter_(!is.na(body_mass_g)) then group_by({{ group }})
# and summarize.) Test it with avg_mass_by(penguins, species).
# C2. (Stretch) Generalise to avg_by(df, group, var) that averages any
# numeric column by any group.
## Part D: functions + simulation ---------------------------------
# D1. Turn your CLT simulation from Lab 2 into a function
# clt_sim <- function(n, R = 1000, dist = "exp") returning the
# vector of R means. Let dist choose between rexp(n) and
# rbinom(n, 1, 0.2). Plot a histogram for each distribution.
# D2. Use stopifnot() to add sanity checks to one of your functions
# (e.g. that the input is numeric). Run it on a character input
# and read the error.
## Part E: STRETCH ---------------------------------
# E1. Write a function that uses any() and all() to report whether a
# vector contains NAs, and if so, how many.
# E2. (replicate) The slide introduced replicate(R, expr) to re-run an
# expression R times. Use it to estimate the sampling distribution
# of the median:
# (a) replicate(1000, median(runif(20))): get the vector of medians,
# then its mean and sd. Compare with the mean of
# replicate(1000, median(runif(200))): a larger sample shrinks
# the spread of the median.
# (b) Wrap the whole thing into a helper sample_medians(n, R = 1000)
# = replicate(R, median(runif(n))) and confirm
# sd(sample_medians(20)) > sd(sample_medians(200)).