不平衡数据集的变量创建

时间:2012-03-05 21:00:33

标签: r

我有一个数据集,其中观察值为ID year event_type event_date。每ID year个观察数量不平衡。具体而言,这些是冲突年代的战争结果。每场战斗都有一个日期和类型(结果)。

我想要做的是根据ID year子集中某种类型的事件数创建变量。所以:

ID

year

event_type == x

的总和

我理解如何使用常规for循环执行此操作,但我知道我应该使用tapply(),因为每个ID有不同的观察数量?

2 个答案:

答案 0 :(得分:2)

library(plyr)
df <-data.frame(ID=sample(11:20,25,replace=T),year=sample(1900:1905,25,replace=T),event_type=sample(c("win","lose"),25,replace=T))

# To see this sample data sorted by ID and year.
arrange(df,ID,year)
  ID year event_type
1  11 1901        win
2  11 1904        win
3  11 1910       lose
4  12 1920       lose
5  13 1900        win
6  13 1905        win
7  13 1906       lose
8  13 1912        win
9  13 1920       lose
10 14 1906        win
11 14 1918       lose
12 14 1920        win
13 15 1909        win
14 15 1919        win
15 16 1916        win
16 16 1920       lose
17 18 1901       lose
18 18 1910       lose
19 18 1912       lose
20 18 1920        win
21 19 1916        win
22 19 1916        win
23 19 1917       lose
24 20 1901       lose
25 20 1914       lose



   result <- ddply(df, .(ID,year,event_type),summarise, event_count=length(event_type))

    >result
   ID year event_type event_count
1  11 1903        win           1
2  11 1905       lose           1
3  12 1903       lose           1
4  12 1905        win           1
5  13 1902        win           1
6  13 1905       lose           1
7  14 1903        win           1
8  15 1901        win           2
9  15 1903       lose           1
10 15 1905        win           1
11 16 1904        win           1
12 17 1904       lose           1
13 18 1900       lose           2
14 18 1900        win           1
15 18 1902       lose           1
16 18 1904        win           1
17 18 1905        win           1
18 19 1901       lose           1
19 19 1902        win           1
20 19 1903       lose           1
21 19 1903        win           1
22 20 1901        win           1
23 20 1904        win           1

让我们说你只想计算胜利而不是亏损,那就像是:

result <- ddply(subset(df,event_type=="win"), .(ID,year,event_type),summarise, event_count=length(event_type))

>result
   ID year event_type event_count
1  11 1903        win           1
2  12 1905        win           1
3  13 1902        win           1
4  14 1903        win           1
5  15 1901        win           2
6  15 1905        win           1
7  16 1904        win           1
8  18 1900        win           1
9  18 1904        win           1
10 18 1905        win           1
11 19 1902        win           1
12 19 1903        win           1
13 20 1901        win           1
14 20 1904        win           1

答案 1 :(得分:2)

如果我正确理解了这个问题,那么:

aggregate(event_type ~ ID + year, subset(df,event_type=="x"), length)