Multiply one dataframe by each row in another dataframe and aggregate result

Question

I have one dataframe where each row contain weights. I want to multiply a second dataframe by each row of the first one and aggregate the results with rowsumS then cumprod. For each line in the first dataframe, I want to save one element as a result.

I have achieved this using a for loop, however this is fairly inefficient, particularly for a dataframe with many rows.

Is there a way to to this without a for loop? maybe using tidyverse.

x=runif(4*6)
x=matrix(x,nrow=4,ncol=6)
x_df=as.data.frame(x)

y=rnorm(3*6)
y=matrix(y,nrow=3,ncol=6)
y_df=as.data.frame(y)

ret=rep(NA, nrow(x_df))

for (i in 1:nrow(x_df)){
  rets=as.data.frame(mapply('*',y_df,x_df[i,]))
  rets=tail(cumprod(1+rowSums(rets,na.rm = TRUE)),1)
  ret[i]=rets
}

the vector rets contain the desired result.

Answer 1

Any of the following would work:

Base R:

apply(tcrossprod(y, x) + 1, 2, prod)
[1]  0.3222529  0.1435537 -0.3998603 -2.1293011

Using matrixStats :

matrixStats::rowProds(tcrossprod(x, y) + 1)
[1]  0.3222529  0.1435537 -0.3998603 -2.1293011

matrixStats::colProds(tcrossprod(y, x) + 1)
[1]  0.3222529  0.1435537 -0.3998603 -2.1293011

---

your code:

set.seed(1)
x=runif(4*6)

x=matrix(x,nrow=4,ncol=6)
x_df=as.data.frame(x)

y=rnorm(3*6)
y=matrix(y,nrow=3,ncol=6)
y_df=as.data.frame(y)

ret=rep(NA, nrow(x_df))

for (i in 1:nrow(x_df)){
  rets=as.data.frame(mapply('*',y_df,x_df[i,]))
  rets=tail(cumprod(1+rowSums(rets,na.rm = TRUE)),1)
  ret[i]=rets
}
ret
[1]  0.3222529  0.1435537 -0.3998603 -2.1293011