在 R 中查找具有最近较高值的列

Question

I'm trying to find the index (or the name) of the column with the closest value to another column. More precisely, I have a dataset that look like:

data <- data.frame(cum_1 = c(1,2), 
           cum_2 = c(2,3),
           cum_3 = c(3,4),
           median = c(1, 2.2))

And I'm trying to come up with a function telling me which of the cum_i column gives me the nearest number above the associated median value. With the dataset provided above for instance, the function would tell me that cum_1 provides it for the first row, and cum_2 does it in the second row. (Or in index notation).

Any help appreciated, thanks!

Answer 1

I'm sure there are more elegant ways to do this but here's a base R start:

apply(data, 1, function(x) {
    val <- x[-length(x)] - x[length(x)]; which.min(replace(val, val < 0, NA))})
#[1] 1 2

Explanations:

1. apply(data, 1, function(x)...) applies the function to every row (since MARGIN = 1 ).
2. The function calculates val for every row as the values from all columns except the last x[-length(x)] minus the value in the last column x[length(x)] (the median column). The function then returns the index of the row vector (ie the index of the column) with the closest but higher value to the median value. We get the "closest but higher" value by replacing all closest but smaller values with NA .

Answer 2

You can use max.col for a vectorised option:

max.col(data[-ncol(data)] - data$median >= 0, ties.method = 'first')
#[1] 1 2

---

data[-ncol(data)] removes the last column ( median ) and subtracts each of these columns with data$median to get:

data[-ncol(data)] - data$median
#  cum_1 cum_2 cum_3
#1   0.0   1.0   2.0
#2  -0.2   0.8   1.8

We compare this output with >= 0 to get TRUE / FALSE values

data[-ncol(data)] - data$median >= 0

#     cum_1 cum_2 cum_3
#[1,]  TRUE  TRUE  TRUE
#[2,] FALSE  TRUE  TRUE

Since TRUE > FALSE we can use max.col to get column index of maximum value in each row. In case, if there are more than 1 value which are TRUE we specify ties.method = 'first to get the first index.