How can I count a variable in R conditional on the value of another variable? - r

Question

I want to count occurrences of a variable in a dataframe by another variable, conditional on the value of a third variable. Here is my data:
Name Store Purchase Date
John CVS Shampoo 1/1/2001
John CVS Toothpaste 1/1/2001
John Whole Foods Kombucha 1/1/2005
John Kroger Ice Cream 1/1/2002
Jane CVS Soap 1/1/2001
Jane Whole Foods Crackers 1/1/2004
For each purchase, I want a count of how many previous purchases were made by the specified person, and how many previous shopping trips, like this:
Name Store Purchase Date Prev_Purchase Prev_trip
John CVS Shampoo 1/1/2001 0 0
John CVS Toothpaste 1/1/2001 0 0
John Whole Foods Kombucha 1/1/2005 3 2
John Kroger Ice Cream 1/1/2002 2 1
Jane CVS Soap 1/1/2001 0 0
Jane Whole Foods Crackers 1/1/2004 1 1
If I wanted the total number of purchases/trips for each person, I would use count or tapply--is there a way to adapt these functions so that the outputs are conditional on a third variable (date)?

Answer 1

Maybe you can try the base R code using ave
transform(df,
Prev_Purchase = ave(as.numeric(as.Date(Date, "%d/%m/%Y")), Name, FUN = function(x) sapply(x, function(p) sum(p > x))),
Prev_trip = ave(as.numeric(as.Date(Date, "%d/%m/%Y")), Name, FUN = function(x) sapply(x, function(p) length(unique(x[p > x]))))
)
which gives
Name Store Purchase Date Prev_Purchase Prev_trip
1 John CVS Shampoo 1/1/2001 0 0
2 John CVS Toothpaste 1/1/2001 0 0
3 John Whole Foods Kombucha 1/1/2005 3 2
4 John Kroger Ice Cream 1/1/2002 2 1
5 Jane CVS Soap 1/1/2001 0 0
6 Jane Whole Foods Crackers 1/1/2004 1 1
Data
df <- structure(list(Name = c("John", "John", "John", "John", "Jane",
"Jane"), Store = c("CVS", "CVS", "Whole Foods", "Kroger", "CVS",
"Whole Foods"), Purchase = c("Shampoo", "Toothpaste", "Kombucha",
"Ice Cream", "Soap", "Crackers"), Date = c("1/1/2001", "1/1/2001",
"1/1/2005", "1/1/2002", "1/1/2001", "1/1/2004")), class = "data.frame", row.names = c(NA,
-6L))

Answer 2

I think it should solve your problem. If your data is huge it's better if you optimize this code chunk.
# load environment
library(lubridate)
# base function
AddInfo = function(name, date, df) {
prev_purchase = sum(df$Name == name & df$Date < date)
prev_trip = length(unique(filter(df, Name == name & Date < date)$Date))
data = data.frame(
Prev_purchase = prev_purchase,
Prev_trip = prev_trip
)
return(data)
}
# define data frame
df = data.frame(
Name = c(rep('John', 4), rep('Jane', 2)),
Store = c('CVS', 'CVS', 'Whole Foods', 'Kroger', 'CVS', 'Whole Foods'),
Purchase = c('Shampoo', 'Toothpaste', 'Kombucha', 'Ice Cream', 'Soap', 'Crackers'),
Date = c('1/1/2001', '1/1/2001', '1/1/2005', '1/1/2002', '1/1/2001', '1/1/2004')
)
# transform date to POSIXct
df$Date = dmy(df$Date)
# apply function and bind the results
cols = mapply(AddInfo, df$Name, df$Date, MoreArgs = list(df), SIMPLIFY = FALSE)
cols = bind_rows(cols)
df = cbind(df, cols)
Here is the output:
Name Store Purchase Date Prev_purchase Prev_trip
1 John CVS Shampoo 1/1/2001 0 0
2 John CVS Toothpaste 1/1/2001 0 0
3 John Whole Foods Kombucha 1/1/2005 3 2
4 John Kroger Ice Cream 1/1/2002 2 1
5 Jane CVS Soap 1/1/2001 0 0
6 Jane Whole Foods Crackers 1/1/2004 1 1

Answer 3

We could also use outer
library(dplyr)
library(lubridate)
df %>%
mutate(Date = dmy(Date)) %>%
group_by(Name) %>%
mutate(Prev_Purchase = colSums(outer(Date, Date, FUN = "<")),
Prev_trip = colSums(outer(unique(Date), Date, FUN = "<")))
# A tibble: 6 x 6
# Groups: Name [2]
# Name Store Purchase Date Prev_Purchase Prev_trip
# <chr> <chr> <chr> <date> <dbl> <dbl>
#1 John CVS Shampoo 2001-01-01 0 0
#2 John CVS Toothpaste 2001-01-01 0 0
#3 John Whole Foods Kombucha 2005-01-01 3 2
#4 John Kroger Ice Cream 2002-01-01 2 1
#5 Jane CVS Soap 2001-01-01 0 0
#6 Jane Whole Foods Crackers 2004-01-01 1 1