Assign different values to a large number of columns - r

Question

I have a large set of financial data that has hundreds of columns. I have cleaned and sorted the data based on date. Here is a simplified example:
df1 <- data.frame(matrix(vector(),ncol=5, nrow = 4))
colnames(df1) <- c("Date","0.4","0.3","0.2","0.1")
df1[1,] <- c("2000-01-31","0","0","0.05","0.07")
df1[2,] <- c("2000-02-29","0","0.13","0.17","0.09")
df1[3,] <- c("2000-03-31","0.03","0.09","0.21","0.01")
df1[4,] <- c("2004-04-30","0.05","0.03","0.19","0.03")
df1
Date 0.4 0.3 0.2 0.1
1 2000-01-31 0 0 0.05 0.07
2 2000-02-29 0 0.13 0.17 0.09
3 2000-03-31 0.03 0.09 0.21 0.01
4 2000-04-30 0.05 0.03 0.19 0.03
I assigned individual weights (based on market value from the raw data) as column headers, because I don’t care about the company names and I need the weights for calculating the result.
My ultimate goal is to get: 1. Sum of the weighted returns; and 2. Sum of the weights when returns are non-zero. With that being said, below is the result I want to get:
Date SWeightedR SWeights
1 2000-01-31 0.017 0.3
2 2000-02-29 0.082 0.6
3 2000-03-31 0.082 1
4 2000-04-30 0.07 1
For instance, the SWeightedR for 2000-01-31 = 0.4x0+0.3x0+0.2x0.05+0.1x0.07, and SWeights = 0.2+0.1.
My initial idea was to assign the weights to each column like WCol2 <- 0.4, then use cbind to create new columns and use c(as.matrix() %*% ) to get the sums. Soon I realize that this is impossible as there are hundreds of columns. Any advice or suggestion is appreciated!

Answer 1

Here's a simple solution using matrix multiplications (as you were suggesting yourself).
First of all, your data seem to be of character type and I'm not sure it's the real case with the real data, but I would first convert it to an appropriate type
df1[-1] <- lapply(df1[-1], type.convert)
Next, we will convert the column names to a numeric class too
vec <- as.numeric(names(df1)[-1])
Finally, we could easily create the new columns in two simple steps. This indeed has a to matrix conversion overhead, but maybe you should work with matrices in the first place. Either way, this is fully vectorized
df1["SWeightedR"] <- as.matrix(df1[, -1]) %*% vec
df1["SWeights"] <- (df1[, -c(1, ncol(df1))] > 0) %*% vec
df1
# Date 0.4 0.3 0.2 0.1 SWeightedR SWeights
# 1 2000-01-31 0.00 0.00 0.05 0.07 0.017 0.3
# 2 2000-02-29 0.00 0.13 0.17 0.09 0.082 0.6
# 3 2000-03-31 0.03 0.09 0.21 0.01 0.082 1.0
# 4 2004-04-30 0.05 0.03 0.19 0.03 0.070 1.0
Or, you could convert to a long format first (here's a data.table example), though I believe it will be less efficient as this are basically by row operations
library(data.table)
res <- melt(setDT(df1), id = 1L, variable.factor = FALSE
)[, c("value", "variable") := .(as.numeric(value), as.numeric(variable))]
res[, .(SWeightedR = sum(variable * value),
SWeights = sum(variable * (value > 0))), by = Date]
# Date SWeightedR SWeights
# 1: 2000-01-31 0.017 0.3
# 2: 2000-02-29 0.082 0.6
# 3: 2000-03-31 0.082 1.0
# 4: 2004-04-30 0.070 1.0