循环并保存不同长度的循环输出[重复]

Question

This question is an exact duplicate of:

• Iterate calculations of different length 1 answer

Seemingly elementary question, but I've searched and tried several options and cannot get the output to appear. I really appreciate the help on the below:

I have a column of returns (dput below) and i'm compounding them in a loop beginning with each element all the way through the end. I'd like to save the output in "results", which I initialized beforehand. I tried several ways of putting the output into results, as seen below, but none are successful. However, I see with print(temp) that the calculations are happening as I'd like them.

Each iteration yields an answer of different length, however, and I want each to be in its own row (which is why I made extra room of NAs in the vector). Using the current method: results=resultstemp is not successful because the "extra" columns for all but the first iteration are filled with repeats of the beginning of the result. So for instance, the second-to-last row has the same two numbers (the full result for a compounding of only two returns) repeat 12 times through the entire row.

    #results=NULL
    results=matrix(data=NA,nrow=nrow(returns),ncol=nrow(returns))
    for (i in 1:nrow(returns)) {
      temp=cumprod(1+returns[i:nrow(returns),]/100)-1
      #resultstemp=append(results,temp)
      #resultstemp[i,]=temp
      resultstemp=rbind(results,temp)
      #print(resultstemp)
      print(temp)
      results=resultstemp
      #resultstemp[i,]=temp
      #results[i,]=resultstemp
      #results[i]=resultstemp[i]
      #results[i]=temp[i]
    }

Returns:

structure(list(`runif(24, 0, 10)` = c(7.60884596500546, 4.31712970370427, 
1.7181678651832, 4.86275367671624, 8.06177110411227, 8.07952171890065, 
7.45263583026826, 9.86292108893394, 4.06634262995794, 2.36454397207126, 
9.12716506049037, 3.72667369898409, 1.2204843852669, 7.80610600719228, 
0.640116988215595, 6.94793848553672, 1.73743493855, 2.57189674302936, 
4.7653386532329, 1.79362375289202, 7.56623527035117, 2.70907687023282, 
4.45359382545575, 5.50409059040248)), .Names = "runif(24, 0, 10)", row.names = c(NA, 
-24L), class = "data.frame")

Answer 1

If I understand you correctly, you want something like a lower triangular matrix, where the first column is just the first percentage, the second is the first compounded with the second, etc. And in each row, you go out one more period? If so, I think the following will work for you:

CMPND <- function(X){
  n <- length(X)
  OUTPUT <- matrix(NA, nrow = n, ncol = n)
  for (i in seq_len(n)){
    OUTPUT[i, 1] = X[1] / 100
    for (j in seq_len(i - 1) + 1) {
      OUTPUT[i, j] = ((1 + OUTPUT[i, j - 1]) * (1 + X[j] / 100)) - 1
    }
  }
  return(OUTPUT)
}

Using your data is long, but give a sequence of 5 10% returns, you get something like:

> CMPND(rep(.1, 5))
      [,1]     [,2]        [,3]        [,4]       [,5]
[1,] 0.001       NA          NA          NA         NA
[2,] 0.001 0.002001          NA          NA         NA
[3,] 0.001 0.002001 0.003003001          NA         NA
[4,] 0.001 0.002001 0.003003001 0.004006004         NA
[5,] 0.001 0.002001 0.003003001 0.004006004 0.00501001

You can make it more efficient using memoization , I'm sure.