Calculating sum of certain values across two columns in R - r

Question

I currently have a dataframe like the one below of a bunch of pairwise correlations:
Data
structure(list(ID1 = c("A", "A", "A", "B", "B", "C"), ID2 = c("B",
"C", "D", "C", "D", "D"), cor = c(0.6, 0.6, 0.2, 0.1, 0.9, 0.2
), value1 = c(50L, 50L, 50L, 20L, 20L, 30L), value2 = c(20L,
30L, 100L, 30L, 100L, 100L)), class = "data.frame", row.names = c(NA,
-6L))
ID1 ID2 cor value1 value2
1 A B 0.6 50 20
2 A C 0.6 50 30
3 A D 0.2 50 100
4 B C 0.1 20 30
5 B D 0.9 20 100
6 C D 0.2 30 100
I'm trying to get the sum of all IDs (i.e. B) of the product between cor and either value1 or value2 depending on whether it is from ID1 or ID2.
For instance, the sum of B would be (cor x value)
(0.6 x 50) + (0.1 x 30) + (0.9 x 100)
I essentially would need to do this for around 20000 unique IDs. I hope this makes sense. I'm not that great in R (yet)!

Answer 1

Does this achieve what you need?
library(tidyverse)
df2 <- df %>%
pivot_longer(names_to = "names", values_to = "values", -c(cor:value2)) %>%
mutate(value = if_else(names == "ID1", value2, value1),
sum = cor * value) %>%
group_by(values) %>%
summarise(sum = sum(sum))

Answer 2

Unless you're looking for dplyr way of answering it, here's a quick but a little inelegant way of doing it:
cond1 <- df$ID1[df$ID1 == "B"]
sum1 <- sum(df$cor[cond1] * df$value1[cond1])
cond2 <- df$ID2[df$ID2 == "B"]
sum2 <- sum(df$cor[cond2] * df$value2[cond2])
finalsum = sum1 + sum2
Basically you want to first look at which row B is in ID1, and then do the product-sum, and then look at which row B is in ID2 and do the same.
Update:
What if you have thousands of ID? Again, I like it quick so create a function out of it:
prodsum <- function (df, ID) {
cond1 <- df$ID1[df$ID1 == ID]
sum1 <- sum(df$cor[cond1] * df$value1[cond1])
cond2 <- df$ID2[df$ID2 == ID]
sum2 <- sum(df$cor[cond2] * df$value2[cond2])
return(sum1 + sum2)
}
Then prodsum(df, "B") will give you the answer for original question. And you can use sapply() to do the job of cycling through thousands of IDs:
IDs <- unique(c(df$ID1, df$ID2))
sapply(IDs, function (x) prodsum(df, x)
There may or may not be a problem if an ID exists exclusively in ID1 or ID2. I'm sure you can write a conditional to deal with the problem.

Answer 3

Another way of looking the thing as below.
Assuming your data frame name is a
a1 <- subset(a,select=c(ID1,cor,value1))
a2 <- subset(a,select=c(ID2,cor,value1))
colnames(a2)[colnames(a2) == "ID2"] <- "ID1"
a3 <- rbind(a1,a2)
a3$MULTIPLY1 <- a3$cor * a3$value1
a4 <- a3 %>% group_by(ID1) %>% summarise(FINALVALUE = sum(MULTIPLY1))
# A tibble: 4 x 2
ID1 FINALVALUE
<chr> <dbl>
1 A 70
2 B 50
3 C 38
4 D 34
Hope this will help to some extent...!