假设给定长度为k的向量x
,我想获得一个k x k的矩阵X
,其中X[i,j]
是x[i]+...+x[j]
的总和。我现在的做法是
set.seed(1)
x <- rnorm(10)
X <- matrix(0,10,10)
for(i in 1:10)
for(j in 1:10)
X[i,j] <- sum(x[i:j])
# [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
# [1,] -0.6264538 -0.4428105 -1.2784391 0.3168417 0.64634948 -0.1741189 0.31331014 1.0516348 1.6274162 1.3220278
# [2,] -0.4428105 0.1836433 -0.6519853 0.9432955 1.27280329 0.4523349 0.93976395 1.6780887 2.2538700 1.9484816
# [3,] -1.2784391 -0.6519853 -0.8356286 0.7596522 1.08915996 0.2686916 0.75612063 1.4944453 2.0702267 1.7648383
# [4,] 0.3168417 0.9432955 0.7596522 1.5952808 1.92478857 1.1043202 1.59174924 2.3300739 2.9058553 2.6004669
# [5,] 0.6463495 1.2728033 1.0891600 1.9247886 0.32950777 -0.4909606 -0.00353156 0.7347931 1.3105745 1.0051861
# [6,] -0.1741189 0.4523349 0.2686916 1.1043202 -0.49096061 -0.8204684 -0.33303933 0.4052854 0.9810667 0.6756783
# [7,] 0.3133101 0.9397640 0.7561206 1.5917492 -0.00353156 -0.3330393 0.48742905 1.2257538 1.8015351 1.4961467
# [8,] 1.0516348 1.6780887 1.4944453 2.3300739 0.73479315 0.4052854 1.22575376 0.7383247 1.3141061 1.0087177
# [9,] 1.6274162 2.2538700 2.0702267 2.9058553 1.31057450 0.9810667 1.80153511 1.3141061 0.5757814 0.2703930
# [10,] 1.3220278 1.9484816 1.7648383 2.6004669 1.00518611 0.6756783 1.49614672 1.0087177 0.2703930 -0.3053884
但我无法摆脱这种感觉,认为必须有一种更优雅的R方式(除了将其转换为Rcpp)。
rollapply()
函数,怎么样? - Gopalafor
循环并不那么糟糕,因为这些循环中没有任何东西“增长”。它们只是填充一个已经具有最终大小的矩阵。如果你在循环中构建一个矩阵,例如通过rbind
,那么for
循环会变得很慢。 - mra68