我是堆栈溢出和R初学者的新手。
我想计算一个大数据集的回报,如下所示:
function initialize() {
function calcRoute() {
var directionsService = new google.maps.DirectionsService();
var start = "52.4076963,-1.4853391999999985";
var end = "52.6114729,-1.6812878000000637";
var request = {
origin: start,
destination: end,
travelMode: google.maps.TravelMode.DRIVING
};
directionsService.route(request, function(response, status) {
if (status == google.maps.DirectionsStatus.OK) {
var directionsDisplay = new google.maps.DirectionsRenderer();
directionsDisplay.setPanel(document.getElementById('result'));
directionsDisplay.setDirections(response);
}
});
}
calcRoute();
}
google.maps.event.addDomListener(window, "load", initialize);
这是一种<script src="https://maps.googleapis.com/maps/api/js"></script>
<div id="result"></div>
格式,比方说价格,列是公司,值是价格,实际数据集有更多列和行。我想建立一个新的DT来计算月收益,我知道您可以使用Date C1 C2 C3
31.01.1985 NA 47 NA
28.02.1985 NA 45 NA
29.03.1985 130 56 NA
30.04.1985 140 67 NA
31.05.1985 150 48 93
28.06.1985 160 79 96
31.07.1985 160 56 94
30.08.1985 160 77 93
30.09.1985 160 66 93
31.10.1985 160 44 93
29.11.1985 160 55 93
函数来做到这一点。但是如何建立具有如此多列而又没有for循环的新数据表?
我想到了:
data.table
但这只是出于某种原因而给出:
diff()
谢谢。
答案 0 :(得分:2)
获得该输出的原因是因为Prices[, names(Prices) != "Date"]
返回了逻辑向量:
> Prices[, names(Prices) != "Date"]
[1] FALSE TRUE TRUE TRUE
由于可以使用逻辑进行计算,因此也可以在逻辑向量上使用diff
。然后将FALSE
视为0
,将TRUE
视为1
。所以基本上您在做diff(c(0,1,1,1))
。
您想要的解决方案:
cols <- setdiff(names(Prices),"Date")
# option 1:
Prices[, paste0(cols,"_return") := lapply(.SD, function(x) (x - shift(x, fill = NA))/shift(x, fill = NA)), .SDcols = cols][]
# option 2:
Prices[, paste0(cols,"_return") := lapply(.SD, function(x) c(NA,diff(x))/shift(x, fill = NA)), .SDcols = cols][]
给出:
> Prices Date C1 C2 C3 C1_return C2_return C3_return 1: 1985-01-31 NA 47 NA NA NA NA 2: 1985-02-28 NA 45 NA NA -0.04255319 NA 3: 1985-03-29 130 56 NA NA 0.24444444 NA 4: 1985-04-30 140 67 NA 0.07692308 0.19642857 NA 5: 1985-05-31 150 48 93 0.07142857 -0.28358209 NA 6: 1985-06-28 160 79 96 0.06666667 0.64583333 0.03225806 7: 1985-07-31 160 56 94 0.00000000 -0.29113924 -0.02083333 8: 1985-08-30 160 77 93 0.00000000 0.37500000 -0.01063830 9: 1985-09-30 160 66 93 0.00000000 -0.14285714 0.00000000 10: 1985-10-31 160 44 93 0.00000000 -0.33333333 0.00000000 11: 1985-11-29 160 55 93 0.00000000 0.25000000 0.00000000
如果要创建新的data.table
,可以使用以下两个选项之一:
# option 1:
Returns <- Prices[, c(list(Date = Date), lapply(.SD, function(x) (x - shift(x, fill = NA))/shift(x, fill = NA))), .SDcols = cols]
# option 2:
Returns <- copy(Prices)
Returns[, (cols) := lapply(.SD, function(x) (x - shift(x, fill = NA))/shift(x, fill = NA)), .SDcols = cols]
使用的数据:
Prices <- fread("Date C1 C2 C3
31.01.1985 NA 47 NA
28.02.1985 NA 45 NA
29.03.1985 130 56 NA
30.04.1985 140 67 NA
31.05.1985 150 48 93
28.06.1985 160 79 96
31.07.1985 160 56 94
30.08.1985 160 77 93
30.09.1985 160 66 93
31.10.1985 160 44 93
29.11.1985 160 55 93")[, Date := as.Date(Date, "%d.%m.%Y")]
答案 1 :(得分:1)
我会写一个函数来处理单列值
pc.change <- function(x) {
(c(x[2:length(x)], NA) - x)*100/x }
然后将其应用于所有值列的矩阵
d <- read.table(text = "Date C1 C2 C3
31.01.1985 NA 47 NA
28.02.1985 NA 45 NA
29.03.1985 130 56 NA
30.04.1985 140 67 NA
31.05.1985 150 48 93
28.06.1985 160 79 96
31.07.1985 160 56 94
30.08.1985 160 77 93
30.09.1985 160 66 93
31.10.1985 160 44 93
29.11.1985 160 55 93", header = TRUE)
apply(as.matrix(d[,2:4]), 2, pc.change)
这给了我
C1 C2 C3
[1,] NA -4.255319 NA
[2,] NA 24.444444 NA
[3,] 7.692308 19.642857 NA
[4,] 7.142857 -28.358209 NA
[5,] 6.666667 64.583333 3.225806
[6,] 0.000000 -29.113924 -2.083333
[7,] 0.000000 37.500000 -1.063830
[8,] 0.000000 -14.285714 0.000000
[9,] 0.000000 -33.333333 0.000000
[10,] 0.000000 25.000000 0.000000
[11,] NA NA NA
然后如果需要,可以将其转换为数据表