How to create a dummy variable corresponding to a change in a value in R-CodePudding

I have the following data:

week <- c(1,2,3,4,1,2,3,4,1,2,3,4)
product <- c("A", "A", "A", "A", "B", "B", "B", "B", "C", "C", "C", "C")
price <- c(5,5,6,6,12,12,11,11, 23, 23, 23, 23)
mydf <- cbind.data.frame(week, product, price)

I want to create a new dummy variable that takes a value of 0 for all initial price levels and 1 if the price increased. The new variable would look as follows:

mydf$price_changed <- c(0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0 ,0)
mydf

CodePudding user response：

We can check with the previous 'price' values (lag) after grouping by 'product' and get the cumulative sum (cumsum)

library(dplyr)
mydf %>%
  group_by(product) %>%
  mutate(price_changed = cumsum(price != lag(price, default = first(price)))) %>% 
  ungroup

-output

# A tibble: 12 × 4
    week product price price_changed
   <dbl> <chr>   <dbl>         <int>
 1     1 A           5             0
 2     2 A           5             0
 3     3 A           6             1
 4     4 A           6             1
 5     1 B          12             0
 6     2 B          12             0
 7     3 B          11             1
 8     4 B          11             1
 9     1 C          23             0
10     2 C          23             0
11     3 C          23             0
12     4 C          23             0

CodePudding user response：

You can use cumany and first:

library(dplyr)
mydf %>% 
  group_by(product) %>% 
  mutate(price_changed =  (cumany(price != first(price))))

# A tibble: 12 × 4
# Groups:   product [3]
    week product price price_changed
   <dbl> <chr>   <dbl>         <int>
 1     1 A           5             0
 2     2 A           5             0
 3     3 A           6             1
 4     4 A           6             1
 5     1 B          12             0
 6     2 B          12             0
 7     3 B          11             1
 8     4 B          11             1
 9     1 C          23             0
10     2 C          23             0
11     3 C          23             0
12     4 C          23             0

or in base R:

with(mydf, ave(price, product, FUN = \(x)  (x != x[1])))

CodePudding user response：

Here is a base R way with ave and cumsum.

week <- c(1,2,3,4,1,2,3,4,1,2,3,4)
product <- c("A", "A", "A", "A", "B", "B", "B", "B", "C", "C", "C", "C")
price <- c(5,5,6,6,12,12,11,11, 23, 23, 23, 23)
mydf <- data.frame(week, product, price)

mydf$price_changed <- with(mydf, ave(price, product, FUN = \(x) cumsum(c(0, x[-length(x)] != x[-1]))))

^{Created on 2022-05-13 by the reprex package (v2.0.1)}

CodePudding user response：

Here is an option with data.table:

library(data.table)

setDT(mydf)[, price_changed := cumsum(price != shift(price, type = "lag", fill = first(price))), by = product]

Output

    week product price price_changed
 1:    1       A     5             0
 2:    2       A     5             0
 3:    3       A     6             1
 4:    4       A     6             1
 5:    1       B    12             0
 6:    2       B    12             0
 7:    3       B    11             1
 8:    4       B    11             1
 9:    1       C    23             0
10:    2       C    23             0
11:    3       C    23             0
12:    4       C    23             0

CodePudding user response：

Option using rleid from data.table:

library(dplyr)
library(data.table)
mydf %>%
  group_by(product) %>%
  mutate(price_changed = rleid(price) - 1)

Output:

# A tibble: 12 × 4
# Groups:   product [3]
    week product price price_changed
   <dbl> <chr>   <dbl>         <dbl>
 1     1 A           5             0
 2     2 A           5             0
 3     3 A           6             1
 4     4 A           6             1
 5     1 B          12             0
 6     2 B          12             0
 7     3 B          11             1
 8     4 B          11             1
 9     1 C          23             0
10     2 C          23             0
11     3 C          23             0
12     4 C          23             0