This might seem whimsical to you, because a similar problem is easily solved in dplyr
. But I still want to know how to do it.
To illustrate, imagine I am looking at employee data and the goal is to find how many records are there for a given employee-date pair.
# Mockup employee data
df <- data.frame(
person_id = c(1, 2, 1),
record_date = as.Date(c("2020-01-01", "2020-01-01", "2020-01-01")),
salary = c(100, 110, 109)
)
# By object counts rows for each unique employee-date pair
out <- by(
data = df,
INDICES = df[, c("win", "record_date")],
FUN = nrow
)
Now the task is to find all those employee-date pairs where the calculated number of rows is more than 1. I couldn’t find answers on the web yet, "by" makes a bad search word. What I can do is something like:
out>1
# record_date
# person_id 2020-01-01
# 1 TRUE
# 2 FALSE
But I am not sure how to get (1, "2020-01-01").
>Solution :
You can use ave
.
transform(df, flag=ave(person_id, person_id, record_date, FUN=\(x) length(x) > 1))
# person_id record_date salary flag
# 1 1 2020-01-01 100 1
# 2 2 2020-01-01 110 0
# 3 1 2020-01-01 109 1
You can also use it in subset
.
subset(df, ave(person_id, person_id, record_date, FUN=\(x) length(x) > 1) == 1)
# person_id record_date salary
# 1 1 2020-01-01 100
# 3 1 2020-01-01 109
Note, that ave
internally uses by
.