# Sum values of combinations within a group

For a analysis I would like to transform data from:

```data <- data.frame(
Customer = c("A", "A", "B", "B", "C", "C", "C"),
Product = c("X", "Y", "X", "Z", "X", "Y", "Z"),
Value = c(10, 15, 5, 10, 20, 5, 10)
)
data
#   Customer Product Value
# 1        A       X    10
# 2        A       Y    15
# 3        B       X     5
# 4        B       Z    10
# 5        C       X    20
# 6        C       Y     5
# 7        C       Z    10
```

To:

```Product Product Sum Value
-------|-------|---------
X      |Y      |50
X      |Z      |45
Y      |Z      |15
```

Basically I want to get the sum of the value for every product combination within a customer. I guess it could work with some help of the reshape package but I cannot get it to work.

Here is one way, in two steps:

1) transform your data into a long data.frame of all pairs within customers. For that, I rely on combn to provide the indices of all possible pairs:

```process.one <- function(x) {
n <- nrow(x)
i <- combn(n, 2)
data.frame(Product1 = x\$Product[i[1, ]],
Product2 = x\$Product[i[2, ]],
Value    = x\$Value[i[1, ]] +
x\$Value[i[2, ]])
}

library(plyr)
long <- ddply(data, "Customer", process.one)
long
#   Customer Product1 Product2 Value
# 1        A        X        Y    25
# 2        B        X        Z    15
# 3        C        X        Y    25
# 4        C        X        Z    30
# 5        C        Y        Z    15
```

2) drop the Customer dimension and aggregate your values:

```aggregate(Value ~ ., long[c("Product1", "Product2", "Value")], sum)
#   Product1 Product2 Value
# 1        X        Y    50
# 2        X        Z    45
# 3        Y        Z    15
```