Sum based of the name of variables in R

Question

Let's say that I have a df contains ID, gender, and several numerical variables, and MAX1, MAX2, and MAX3, where

MAX1 = the variable name of the first max values of x1,x2,x3,x4,x5

MAX2 = the variable name of the second max values of x1,x2,x3,x4,x5

MAX3 = the variable name of the third max values of x1,x2,x3,x4,x5

### Generate data
set.seed(123)
    ID <- c(1,2,3,4,5,6,7,8,9,10)
gender <- c("m", "m", "m", "f", "f", "m", "m", "f", "f", "m")
x1 <- rnorm(10, 0, 1)
x2 <- rnorm(10, 0, 1)
x3 <- rnorm(10, 0, 1)
x4 <- rnorm(10, 0, 1)
x5 <- rnorm(10, 0, 1)
df <- data.frame(ID, gender, x1, x2, x3, x4, x5)

maxes <- t(sapply(1:nrow(df), function(i) {
    names(sort(df[i,3:7], decreasing=T)[1:3])
}))
colnames(maxes) <- c("MAX1","MAX2", "MAX3")
df <- cbind(df, maxes)

Now I need to create a new column (call ir m_sum) that has the sum values of MAX1 and MAX2.

For example, for ID=1, MAX1 = x2 and MAX2 = x4, then m_sum the shold be equal to 1.2240818 + 0.42646422 = 1.650546.

Answer 1

How about using apply to do it all in one call?

df$m_sum <- apply(df, 1, function(x) as.double(x[x[ "MAX1" ]]) + as.double(x[x[ "MAX2" ]]))
 #[1] 1.65054602 0.15189652 2.45383397 3.04708946 2.02954308 3.50197809 1.39170465 0.09146139 1.48132102
#[10] 1.17044583

Answer 2

This is complicated by df$MAX1 etc. using factors

but a simple loop something like

sumMAX1MAX2 <- numeric()
for (r in 1:nrow(df)){ 
    sumMAX1MAX2[r] <- df[r, as.character(df$MAX1)[r]] + 
                      df[r, as.character(df$MAX2)[r]]
    }

seems to produce

> sumMAX1MAX2
 [1] 1.65054602 0.15189655 2.45383398 3.04708945 2.02954308 3.50197812
 [7] 1.39170470 0.09146141 1.48132102 1.17044585

Another way would be to do something similar to your maxes, such as

altsumMAX1MAX2 <- sapply(1:nrow(df), function(i){
    sum(sort(df[i,3:7], decreasing=T)[1:2])
    })