我知道R中的filter()
函数计算移动平均线。我想知道是否存在一个函数,它返回移动方差或标准偏差,以便与filter()
函数的输出并排显示在一个图中。
答案 0 :(得分:31)
考虑 zoo 包。例如filter()
给出:
> filter(1:100, rep(1/3,3))
Time Series:
Start = 1
End = 100
Frequency = 1
[1] NA 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
[26] 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
[51] 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
[76] 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 NA
而 zoo 中的rollmean()
给出:
> rollmean(1:100, k = 3, na.pad = TRUE)
[1] NA 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
[26] 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
[51] 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
[76] 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 NA
是相同的(对于本例中的3点移动平均线)。
虽然动物园没有rollsd()
或rollvar()
,但它确实有rollapply()
,其效果与apply()
函数相同R函数到指定的窗口。
> rollapply(1:100, width = 3, FUN = sd, na.pad = TRUE)
[1] NA 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[26] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[51] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
[76] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 NA
Warning message:
In rollapply.zoo(zoo(data), ...) : na.pad argument is deprecated
或更有趣的事情:
> rollapply(vec, width = 3, FUN = sd, na.pad = TRUE)
[1] NA 0.3655067 0.8472871 0.5660495 0.3491970 0.4732417 0.9236859
[8] 0.8075226 1.8725851 1.1930784 0.6329325 1.1412416 0.8430772 0.5808005
[15] 0.3838545 1.1738170 1.1655400 1.3241700 0.6876834 0.1534157 0.4858477
[22] 0.9843506 0.6002713 0.6897541 2.0619563 2.5675788 6.3522039 6.0066864
[29] 6.2618432 5.1704866 2.1360853 2.5602557 1.0408528 1.0316396 4.9441628
[36] 5.0319314 5.7589716 3.2425000 4.8788158 2.0847286 4.5199291 2.5323486
[43] 2.1987149 1.8393000 1.2278639 1.5998965 1.5341485 4.4287108 4.4159166
[50] 4.3224546 3.6959067 4.9826264 5.3134044 8.4084322 9.1249234 7.5506725
[57] 3.8499136 3.9680487 5.6362296 4.9124095 4.3452706 4.0227141 4.5867559
[64] 4.7350394 4.3203807 4.4506799 7.2759499 7.6536424 7.8487654 2.0905576
[71] 4.0056880 5.6209853 1.5551659 1.3615268 2.8469458 2.8323588 1.9848578
[78] 1.1201124 1.4248380 1.7802571 1.4281773 2.5481935 1.8554451 1.0925410
[85] 2.1823722 2.2788755 2.4205378 2.0733741 0.7462248 1.3873578 1.4265948
[92] 0.7212619 0.7425993 1.0696432 2.4520585 3.0555819 3.1000885 1.0945292
[99] 0.3726928 NA
Warning message:
In rollapply.zoo(zoo(data), ...) : na.pad argument is deprecated
您可以使用fill = NA
参数删除警告,如
> rollapply(vec, width = 3, FUN = sd, fill = NA)
答案 1 :(得分:19)
TTR套餐包括runSD
:
> library(TTR)
> ls("package:TTR", pattern="run*")
[1] "runCor" "runCov" "runMAD" "runMax" "runMean"
[6] "runMedian" "runMin" "runSD" "runSum" "runVar"
runSD
将比rollapply
快得多,因为它避免了许多R函数调用。例如:
rzoo <- function(x,n) rollapplyr(x, n, sd, fill=NA)
rttr <- function(x,n) runSD(x, n)
library(rbenchmark)
set.seed(21)
x <- rnorm(1e4)
all.equal(rzoo(x,250), rttr(x,250))
# [1] TRUE
benchmark(rzoo(x,250), rttr(x,250))[,1:6]
# test replications elapsed relative user.self sys.self
# 2 rttr(x, 250) 100 0.58 1.000 0.58 0.00
# 1 rzoo(x, 250) 100 54.53 94.017 53.85 0.06
答案 2 :(得分:14)
rollapply
包中的 zoo
采用任意函数。它与filter
的不同之处在于它默认排除了NA
。
尽管如此,加载一个函数包并没有多大意义,因为这个函数很容易让自己滚动(双关语)。
这是一个正确的对齐方式:
my.rollapply <- function(vec, width, FUN)
sapply(seq_along(vec),
function(i) if (i < width) NA else FUN(vec[i:(i-width+1)]))
set.seed(1)
vec <- sample(1:50, 50)
my.rollapply(vec, 3, sd)
[1] NA NA 7.094599 12.124356 16.522712 18.502252 18.193405 7.234178 8.144528
[10] 14.468356 12.489996 3.055050 20.808652 19.467922 18.009257 18.248288 15.695010 7.505553
[19] 10.066446 11.846237 17.156146 6.557439 5.291503 23.629078 22.590558 21.197484 22.810816
[28] 24.433583 19.502137 16.165808 11.503623 12.288206 9.539392 13.051181 13.527749 19.974984
[37] 19.756855 17.616280 19.347696 18.248288 15.176737 6.082763 10.000000 10.016653 4.509250
[46] 2.645751 1.527525 5.291503 10.598742 6.557439
# rollapply output for comparison
rollapply(vec, width=3, sd, fill=NA, align='right')
[1] NA NA 7.094599 12.124356 16.522712 18.502252 18.193405 7.234178 8.144528
[10] 14.468356 12.489996 3.055050 20.808652 19.467922 18.009257 18.248288 15.695010 7.505553
[19] 10.066446 11.846237 17.156146 6.557439 5.291503 23.629078 22.590558 21.197484 22.810816
[28] 24.433583 19.502137 16.165808 11.503623 12.288206 9.539392 13.051181 13.527749 19.974984
[37] 19.756855 17.616280 19.347696 18.248288 15.176737 6.082763 10.000000 10.016653 4.509250
[46] 2.645751 1.527525 5.291503 10.598742 6.557439
答案 3 :(得分:0)
runner
函数在运行的窗口上应用任何R函数。使用runner
可以通过设置长度k
或lag
来指定简单窗口。根据OOP在4元素窗口上的建议,将sd
移动到下方。
library(runner)
set.seed(1)
x <- rnorm(20, sd = 1)
runner(x, sd, k = 4, na_pad = TRUE)
#[1] NA NA NA 1.1021597 0.9967429 1.1556947 0.9884053 0.6902835 0.7180483 0.4647160
#[11] 0.7454670 0.7489618 0.9449882 1.5821988 1.4459037 1.3889432 1.3954101 0.6193867 0.5296744 0.4266423
要在日期窗口上应用运行功能,应指定idx
。 idx
的长度应与x
的长度相同,并且应为日期或整数类型。以下示例说明了大小k = 4
被lag = 1
滞后的窗口。在括号中,每个窗口的索引范围。
idx <- c(4, 6, 7, 13, 17, 18, 18, 21, 27, 31, 37, 42, 44, 47, 48)
runner::runner(x = 1:15,
k = 5,
lag = 1,
idx = idx,
f = function(x) mean(x))
# [1] NA 1.0 1.5 NA 4.0 4.5 4.5 6.0 NA 9.0 NA 11.0 12.0 12.5 13.5