我有一个包含10个变量的简单时间序列数据集-我想创建一个for循环(或函数),为其中的每个变量创建一个“自上个月的变化”变量和一个“自上个月的变化百分比”时间序列(日期除外)。我知道我可以为每个特定的列简单地编写代码,但是由于有很多列,所以我想对其进行优化。
这是我的数据,“日期”,“销售”,“价格”是一些列名称:
+----+---+---+---+---+---+---+---+--
| Date | Sales | Price |
+----+---+---+---+---+---+---+---+--
| 01Aug2019 | 4 | 15 |
| 01Sept2019 | 6 | 30 |
| 01Oct2019 | 10 | 44 |
+----+---+---+---+---+---+---+---+--
这是我希望使用for循环(或任何函数)的样子
+----+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
| Date | Sales | chg_Sales | pct_chg_Sales | Price | chg_Price | pct_chg_Price|
+----+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
| 01Aug2019 | 4 | NA |NA | 15 | NA |NA |
| 01Sept2019 | 6 | 2 |50% | 30 | 15 |100% |
| 01Oct2019 | 10 | 4 |66% | 44 | 14 |46% |
+----+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
我尝试了下面的代码,但是没有用
add_column <- function (x, y){
setDT (x)[,pct_chg_y:= (y - shift (y,1, type="lag")/shift (,1, type="lag")*100]
}
答案 0 :(得分:2)
这里是data.table
的一个选项,其中我们在.SDcols
中指定感兴趣的列,通过从中减去.SD
(Data.table的子集)来创建“ chg_”列。 lag
,即{{1}的shift
,然后在第二步中,使用.SD
将shift
除以'chg_'列来创建'pct_chg / p>
Map
nm1 <- c("Sales", "Price")
setDT(df1)[, paste0("chg_", nm1) := .SD - shift(.SD), .SDcols = nm1]
df1[, paste0("pct_chg_", nm1) :=
Map(function(x, y) 100 * (y/shift(x)), .SD, mget(paste0("chg_", nm1))),
.SDcols = nm1]
df1
# Date Sales Price chg_Sales chg_Price pct_chg_Sales pct_chg_Price
#1: 01Aug2019 4 15 NA NA NA NA
#2: 01Sept2019 6 30 2 15 50.00000 100.00000
#3: 01Oct2019 10 44 4 14 66.66667 46.66667
答案 1 :(得分:1)
library(dplyr)
library(scales)
df1 %>%
arrange(Date) %>%
mutate_at(.vars = c("Sales", "Price"), list(chg = ~(. - lag(.)),
pct_chg = ~percent((. - lag(.))/lag(.))))
# Date Sales Price Sales_chg Price_chg Sales_pct_chg Price_pct_chg
# 1 2019-08-01 4 15 NA NA NA% NA%
# 2 2019-09-01 6 30 2 15 50.0% 100.0%
# 3 2019-10-01 10 44 4 14 66.7% 46.7%