在R中重塑数据矩阵

时间:2014-03-12 14:53:39

标签: r casting dataframe reshape2

我有一些数据要重塑 R ,但无法弄清楚如何。这是一个场景:我有来自不同学校的一些学生的考试成绩数据。以下是一些示例数据:

#Create example data: 
test <- data.frame("score" = c(1,10,20,40,20), "schoolid" = c(1,1,2,2,3))

产生如下数据格式:

  score schoolid
    1        1
   10        1
   20        2
   40        2
   20        3

因此,有学校ID识别学校,并且每个学生都有一个考试分数。对于不同程序的分析,我希望以这样的格式获取数据:

                Score student 1    Score student 2 
School ID == 1        1                   10               
School ID == 2       10                   40
School ID == 3       20                   NA

为了重塑数据,我尝试使用reshape2库中的reshape和cast函数,但这导致了错误:

#Reshape function
reshape(test, v.names = test2$score, idvar = test2$schoolid, direction = "wide")
reshape(test, idvar = test$schoolid, direction = "wide")
#Error: in [.data.frame'(data,,idvar): undefined columns selected

#Cast function
cast(test,test$schoolid~test$score)
#Error: Error: could not find function "cast" (although ?cast works fine)

我猜每个学校的考试成绩数量不同,这一事实使重组过程变得复杂。

我如何重塑这些数据以及我应该使用哪种功能?

2 个答案:

答案 0 :(得分:4)

以下是一些仅使用R的基础的解决方案。所有三个解决方案都使用这个新的studentno变量:

studentno <- with(test, ave(schoolid, schoolid, FUN = seq_along))

1)tapply

with(test, tapply(score, list(schoolid, studentno), c))

,并提供:

   1  2
1  1 10
2 20 40
3 20 NA

2)重塑

# rename score to student and append studentno column
test2 <- transform(test, student = score, score = NULL, studentno = studentno)
reshape(test2, dir = "wide", idvar = "schoolid", timevar = "studentno")

,并提供:

  schoolid student.1 student.2
1        1         1        10
3        2        20        40
5        3        20        NA

3)如果没有得分为0的学生,xtabs xtabs也会有效。

xt <- xtabs(score ~ schoolid + studentno, test)
xt[xt == 0] <- NA  # omit this step if its ok to use 0 in place of NA
xt

,并提供:

        studentno
schoolid  1  2
       1  1 10
       2 20 40
       3 20   

答案 1 :(得分:2)

您必须在某处定义学生ID,例如:

test <- data.frame("score" = c(1,10,20,40,20), "schoolid" = c(1,1,2,2,3))
test$studentid <- c(1,2,1,2,1)

library(reshape2)
dcast(test, schoolid ~ studentid, value.var="score",mean)
  schoolid  1   2
1        1  1  10
2        2 20  40
3        3 20 NaN