每年平均前 5 个数据点

Question

我试图平均每“年”“var1”的前 5 个数据点。我的数据如下。每年数据的长度是不一样的。非常感谢你的帮助！ :)

    year    var1
    2010    1
    2010    2
    2010    3
    2010    4
    2010    2
    2010    4
    2009    2
    2009    3
    2009    4
    2009    1
    2009    3
    2009    4
    2009    2
    2009    5
    2009    3
    2009    6
    2008    4
    2008    2
    2008    3
    2008    4
    2008    1
    2008    3

Answer 1

这样的？

t <- read.csv("t.txt", sep="") ## Read data
myMean <- function(x) ifelse(length(x)<5, mean(x), mean(x[1:5]))
ans <- aggregate(var1 ~ year, data = t, FUN = myMean)
ans
  year var1
1 2008   14
2 2009   13
3 2010   12

我们创建一个函数 myMean 来计算给定向量的前 5 个元素的平均值。 对于ifelse，以防万一，如果有些年份没有 5 个数据点，那么我们取所有数据点的平均值。

我们使用函数aggregate对year的数据集进行分区。对于每个year，我们在var1 上应用函数myMean。

Answer 2

使用data.table，我们将'data.frame'转换为'data.table'（setDT(df1)），按'year'分组，我们得到'var1'的前5个值（head）并获取mean

library(data.table)
setDT(df1)[, list(var1=mean(head(var1,5))), year]
#   year var1
#1: 2010  2.4
#2: 2009  2.6
#3: 2008  2.8

Answer 3

这里还有一个选项，使用split 和lapply：

sapply( split(X$var1,X$year), function(x) ifelse(length(x)<5, mean(x), mean(x[1:5])) )

X 是给定的数据框。

速度对比：

library(microbenchmark)
library(data.table)

X <- read.table( header = TRUE,
                 text = "year    var1
                    2011    3
                    2011    8
                    2010    1
                    2010    2
                    2010    3
                    2010    4
                    2010    2
                    2010    4
                    2009    2
                    2009    3
                    2009    4
                    2009    1
                    2009    3
                    2009    4
                    2009    2
                    2009    5
                    2009    3
                    2009    6
                    2008    4
                    2008    2
                    2008    3
                    2008    4
                    2008    1
                    2008    3" )

myMean <- function(x) ifelse(length(x)<5, mean(x), mean(x[1:5]))

microbenchmark(
  akrun    = setDT(X)[, list(var1=mean(head(var1,5))), year],
  PoChoi.1 = aggregate(var1 ~ year, data = X, FUN = myMean),
  PoChoi.2 = aggregate(var1 ~ year, data = X, FUN = function(x) ifelse(length(x)<5, mean(x), mean(x[1:5]))),
  mra68.1  = sapply( split(X$var1,X$year), myMean ),
  mra68.2  = sapply( split(X$var1,X$year), function(x) ifelse(length(x)<5, mean(x), mean(x[1:5])) ),
  times = 1000
)

# Unit: microseconds
#    expr      min        lq      mean   median       uq        max neval
#    akrun 1781.673 3571.9520 3747.1294 3772.582 3931.320 147014.295  1000
# PoChoi.1 2273.966 4563.1035 4498.3682 4739.505 4982.933   9535.620  1000
# PoChoi.2 2289.817 4571.2555 4515.6098 4733.391 4956.892  21497.376  1000
#  mra68.1  347.368  693.8295  711.6527  731.420  769.462   5615.848  1000
#  mra68.2  346.915  694.7350  717.4941  730.740  772.633   5560.143  1000

akrun 是一个数据表，PoChoi 一个数据框，mra68 一个命名向量：

> akrun
   year var1
1: 2011  5.5
2: 2010  2.4
3: 2009  2.6
4: 2008  2.8
> PoChoi.1
  year var1
1 2008  2.8
2 2009  2.6
3 2010  2.4
4 2011  5.5
> PoChoi.2
  year var1
1 2008  2.8
2 2009  2.6
3 2010  2.4
4 2011  5.5
> mra68.1
2008 2009 2010 2011 
 2.8  2.6  2.4  5.5 
> mra68.2
2008 2009 2010 2011 
 2.8  2.6  2.4  5.5 

一个更大的例子：

library(microbenchmark)
library(data.table)

set.seed(1)

X <- data.frame( year = sample( 1500:2015, 10000, replace=TRUE ),
                 var1 = sample( 1:10, 10000, replace=TRUE ) )

myMean <- function(x) ifelse(length(x)<5, mean(x), mean(x[1:5]))

microbenchmark(
  akrun    = setDT(X)[, list(var1=mean(head(var1,5))), year],
  PoChoi.1 = aggregate(var1 ~ year, data = X, FUN = myMean),
  PoChoi.2 = aggregate(var1 ~ year, data = X, FUN = function(x) ifelse(length(x)<5, mean(x), mean(x[1:5]))),
  mra68.1  = sapply( split(X$var1,X$year), myMean ),
  mra68.2  = sapply( split(X$var1,X$year), function(x) ifelse(length(x)<5, mean(x), mean(x[1:5])) ),
  times = 1000
)

# Unit: milliseconds
# expr      min       lq     mean   median       uq       max neval
# akrun    15.44811 23.50436 36.81674 43.12405 44.22435  69.62202  1000
# PoChoi.1 33.96411 51.52858 83.29682 95.53486 99.60884 241.59967  1000
# PoChoi.2 33.64844 51.70747 83.47835 95.07223 99.44127 247.55881  1000
# mra68.1  11.05145 17.33191 27.21526 31.41954 32.34819 126.89461  1000
# mra68.2  11.05054 17.16615 26.96236 31.25061 32.14054  85.44422  1000

除了不同的类（data.table、data.frame、named vector），结果是相等的：

> all( PoChoi.1 == PoChoi.2 )
[1] TRUE

> all( PoChoi.1$var1 == mra68.1 )
[1] TRUE

> all( mra68.1 == mra68.2 )
[1] TRUE

> all( akrun$var1[order(akrun$year)] == mra68.1 )
[1] TRUE
>