Group rows and add sum column of unique values - r

Question

Here an example of my data.frame:
df = read.table(text='colA colB colC
10 11 7
10 34 7
10 89 7
10 21 7
2 23 5
2 21 5
2 56 5
22 14 3
22 19 3
22 90 3
11 19 2
11 45 2
1 45 0
1 23 0
9 8 0
9 11 0
9 21 0', header = TRUE)
I need to group the rows by colA and colC and add a new column which states the sum of unique values based on colB.
In steps here what I need to do for this specific data.frame:
group rows with colA = 10 and 9, colA = 2 and 1, colA = 22 and colA = 11;
find the unique values of colB per each group;
add the unique values in a new col (newcolD).
Note that colC states the total number of observations for colA = 10 and 9, colA = 2 and 1, colA = 22 and colA = 11.
The data.frame needs to remain ordered decreasingly by colC.
My expected output is:
colA colB colC newcolD
10 11 7 5
10 34 7 5
10 89 7 5
10 21 7 5
9 8 0 5
9 11 0 5
9 21 0 5
2 23 5 4
2 21 5 4
2 56 5 4
1 45 0 4
1 23 0 4
22 14 3 3
22 19 3 3
22 90 3 3
11 19 2 2
11 45 2 2
To note that in df the colB duplicated values are: 11 and 21 for group 10 and 9, and 23 for group 2 and 1.

Answer 1

You can do that with dplyr. The trick is to create a new grouping column which groups consecutive values in colA. This is done with cumsum(c(1, diff(colA) < -1) in the example below.
df1 = read.table(text='colA colB colC
10 11 7
10 34 7
10 89 7
10 21 7
2 23 5
2 21 5
2 56 5
22 14 3
22 19 3
22 90 3
1 45 0
1 23 0
9 8 0
9 11 0
9 21 0', header = TRUE,stringsAsFactors=FALSE)
library(dplyr)
df1 %>%
arrange(desc(colA)) %>%
group_by(group_sequential = cumsum(c(1, diff(colA) < -1))) %>%
mutate(newcolD=n_distinct(colB))
colA colB colC group_sequential newcolD
<int> <int> <int> <dbl> <int>
1 22 14 3 1 3
2 22 19 3 1 3
3 22 90 3 1 3
4 10 11 7 2 5
5 10 34 7 2 5
6 10 89 7 2 5
7 10 21 7 2 5
8 9 8 0 2 5
9 9 11 0 2 5
10 9 21 0 2 5
11 2 23 5 3 4
12 2 21 5 3 4
13 2 56 5 3 4
14 1 45 0 3 4
15 1 23 0 3 4
EDIT FOR NEW DATA
With the data you added, we need to create a custom grouping. I use case_when in the example below. This matches the order you show in the desired output column. In the text, you wrote that you wanted the table to be sorted by colC. To do so, change the last line to arrange(desc(colC))
df1 = read.table(text='colA colB colC
10 11 7
10 34 7
10 89 7
10 21 7
2 23 5
2 21 5
2 56 5
22 14 3
22 19 3
22 90 3
11 19 2
11 45 2
1 45 0
1 23 0
9 8 0
9 11 0
9 21 0', header = TRUE,stringsAsFactors=FALSE)
library(dplyr)
df1 %>%
group_by(group_sequential = case_when(.$colA==10|.$colA==9~1,
.$colA==2|.$colA==1~2,
.$colA==22~3,
.$colA==11~4)) %>%
mutate(newcolD=n_distinct(colB)) %>%
arrange(desc(newcolD))
colA colB colC group_sequential newcolD
<int> <int> <int> <dbl> <int>
1 10 11 7 1 5
2 10 34 7 1 5
3 10 89 7 1 5
4 10 21 7 1 5
5 9 8 0 1 5
6 9 11 0 1 5
7 9 21 0 1 5
8 2 23 5 2 4
9 2 21 5 2 4
10 2 56 5 2 4
11 1 45 0 2 4
12 1 23 0 2 4
13 22 14 3 3 3
14 22 19 3 3 3
15 22 90 3 3 3
16 11 19 2 4 2
17 11 45 2 4 2

Answer 2

You're really not making it easy for us, reposting slight variations of the same question instead of updating the old one and presenting conditions that are vague and inconsistent with what the desired output implies. Anyhow, here is my attempt. This is more an answer to the second question you posted, as that was a bit more general in form.
It's a bit messy, it's pretty much a direct translation of your conditions into a for loop with some if statements. I chose to focus on your written conditions rather than the expected output as that was the easier one to understand. If you want a better answer, please consider cleaning up you question(s) considerably.
df1 <- read.table(text="
colA colB colC
10 11 7
10 34 7
10 89 7
10 21 7
2 23 5
2 21 5
2 56 5
22 14 3
22 19 3
22 90 3
11 19 2
11 45 2
1 45 0
1 23 0
9 8 0
9 11 0
9 21 0", header=TRUE)
df2 <- read.table(text="
colA colB colC
10 11 7
10 34 7
10 89 7
10 21 7
2 23 5
2 21 5
2 56 5
33 24 3
33 78 3
22 14 3
22 19 3
22 90 3
11 19 2
11 45 2
1 45 0
1 23 0
9 8 0
9 11 0
9 21 0
32 11 0", header=TRUE)
df <- df1
for (i in 1:nrow(df)) {
df$colD[i] <- ifelse(df$colC[i] == 0,
0,
length(unique(df$colA[1:i])))
if (any(df$colA[i]-1 == df$colA[1:i]) & df$colC[i] != 0) {
df$colD[i] <- df$colD[which(df$colA[i]-1 == df$colA[1:i])][1]
}
}
# colA colB colC colD
# 10 11 7 1
# 10 34 7 1
# 10 89 7 1
# 10 21 7 1
# 2 23 5 2
# 2 21 5 2
# 2 56 5 2
# 22 14 3 3
# 22 19 3 3
# 22 90 3 3
# 11 19 2 1
# 11 45 2 1
# 1 45 0 0
# 1 23 0 0
# 9 8 0 0
# 9 11 0 0
# 9 21 0 0
df <- df2
for (i in 1:nrow(df)) {
df$colD[i] <- ifelse(df$colC[i] == 0,
0,
length(unique(df$colA[1:i])))
if (any(df$colA[i]-1 == df$colA[1:i]) & df$colC[i] != 0) {
df$colD[i] <- df$colD[which(df$colA[i]-1 == df$colA[1:i])][1]
}
}
df
# colA colB colC colD
# 10 11 7 1
# 10 34 7 1
# 10 89 7 1
# 10 21 7 1
# 2 23 5 2
# 2 21 5 2
# 2 56 5 2
# 33 24 3 3
# 33 78 3 3
# 22 14 3 4
# 22 19 3 4
# 22 90 3 4
# 11 19 2 1
# 11 45 2 1
# 1 45 0 0
# 1 23 0 0
# 9 8 0 0
# 9 11 0 0
# 9 21 0 0
# 32 11 0 0
To also group the rows where colC is zero, it's sufficient to adjust the conditionals like this:
for (i in 1:nrow(df)) {
df$colD[i] <- length(unique(df$colA[1:i]))
if (any(df$colA[i]-1 == df$colA[1:i])) {
df$colD[i] <- df$colD[which(df$colA[i]-1 == df$colA[1:i])][1]
}
}