Using mutate and group_by to roll an operation over rows - r

Question

I have the following data:
country year sales
--------------------------
Afghanistan 1950 30
Afghanistan 1951 35
Albania 1950 0
Albania 1951 5
total 1950 30
total 1951 40
I want to generate a new column, ratio, which is the ratio of sales for any given country-year combination to the total for that year. So the output should be:
country year sales ratio
---------------------------------
Afghanistan 1950 30 1
Afghanistan 1951 35 0.875
Albania 1950 0 0
Albania 1951 5 0.125
total 1950 30 1
total 1951 40 1
I'd like to use tidyverse (which I am somewhat new to) to accomplish this, but I'm still somewhat confused about how to use mutate and group_by to accomplish this (or even if that is the best way to go about this task in general).
I tried unsuccessfully to use the advice given in this thread. What I have tried is:
library(tidyverse)
df <- df %>%
group_by(year) %>%
mutate(ratio = sales[country]/sales[country == "total"])
But this generates a column called ratio full of NAs. Do I need to use a loop or something else? I'm somewhat new to R and I will admit I have avoided loops up until now. Looking over documentation on loops, I couldn't quite think of how I would use one to run over each country-year combination and generate a new column.

Answer 1

You can group by country and then divide sales by maximum of sales - which is total, I suppose.
library(dplyr)
df %>%
group_by(year) %>%
mutate(ratio = sales / max(sales))
# A tibble: 6 x 4
# Groups: year [2]
# country year sales ratio
# <chr> <int> <int> <dbl>
#1 Afghanistan 1950 30 1
#2 Afghanistan 1951 35 0.875
#3 Albania 1950 0 0
#4 Albania 1951 5 0.125
#5 total 1950 30 1
#6 total 1951 40 1
In base R
transform(df, ratio = ave(sales, year, FUN = function(x) x / max(x)))
Or with data.table
library(data.table)
setDT(df)[, ratio := sales / max(sales), by = year][]
data
df <- structure(list(country = c("Afghanistan", "Afghanistan", "Albania",
"Albania", "total", "total"), year = c(1950L, 1951L, 1950L, 1951L,
1950L, 1951L), sales = c(30L, 35L, 0L, 5L, 30L, 40L)), .Names = c("country",
"year", "sales"), class = "data.frame", row.names = c(NA, -6L
))