如何按行和列>列名查找？

Question

我正在考虑如何按大学名称（第一行：A,...,F）、字段名称（第一列：Acute,.. .,En) 和/或毕业时间 (time) 在以下文件 DS.csv 中。 我在想dplyr 方法，但无法将数字 ID 查找（线程答案How to overload function parameters in R?）扩展到三个变量的查找。 挑战

1. 如何按第一行查找？也许，类似于$1 == "A"。
2. 如何将大学查询扩展到两列？伪代码$1 == "A" 大约是第二列和第三列，...，$1 == "F" 大约是最后两列。

3. 通过 3 个查找条件进行查找：第一行（无标题）、带有标题 Field 的第一列和标题 time。伪代码

   times <- getTimes($1 == "A", Field == "Ane", by = "desc(time)")
   

数据DS.csv 有数据。第一列表示实验。以下数据采用交叉表格式，因此

,A,,B,,C,,D,,E,,F,
Field,time,T,time,T,time,T,time,T,time,T,time,T
Acute,0,0,8.3,1,7.5,1,8.6,2,0,0,8.3,4
Ane,9,120,7.7,26,7.9,43,7.8,77,7.9,60,8.2,326
En,15.6,2,12.9,1,0,0,0,0,14.3,1,14.6,4
Fo,9.2,2,0,0,5.4,1,0,0,0,0,7.9,3

并采用直表格式，这样

Field,time,T,Experiment       
Acut,0,0,A
An,9,120,A
En,15.6,2,A
Fo,9.2,2,A
Acute,8.3,1,B       
An,7.7,26,B
En,12.9,1,B
Fo,0,0,B  
Acute,7.5,1,C       
An,7.9,43,C
En,0,0,C  
Fo,5.4,1,C
Acute,8.6,2,D       
An,7.8,77,D
En,0,0,D  
Fo,0,0,D  
Acute,0,0,E         
An,7.9,60,E
En,14.3,1,E
Fo,0,0,E  
Acute,8.3,4,F       
An,8.2,326,F
En,14.6,4,F
Fo,7.9,3,F

伪代码

library('dplyr')
ow <- options("warn")
DF <- read.csv("/home/masi/CSV/DS.csv", header = T)

# Lookup by first row, Lookup by Field, lookup by Field's first column?
times <- getTimes($1 == "A", Field == "Ane", by = "desc(time)")

预期输出：9
广义预期输出：a, b, c, ...

## Data where values marked by small letters a,b,c, ... are wanted 
#       uni1    uni2 ... 
#       time T  time T ...
#Field1 a       c
#Field2 b       ...
#...    ...

R：3.3.3 (2017-03-06)
操作系统：Debian 8.7
硬件：华硕 Zenbook UX303UA

Answer 1

以您的初始原始数据为起点：

# read the data & skip 1st & 2nd line which contain only header information
DF <- read.csv(text=",A,,B,,C,,D,,E,,F,
Field,time,T,time,T,time,T,time,T,time,T,time,T
Acute,0,0,8.3,1,7.5,1,8.6,2,0,0,8.3,4
Ane,9,120,7.7,26,7.9,43,7.8,77,7.9,60,8.2,326
En,15.6,2,12.9,1,0,0,0,0,14.3,1,14.6,4
Fo,9.2,2,0,0,5.4,1,0,0,0,0,7.9,3", header=FALSE, stringsAsFactors=FALSE, skip=2)

# read the first two lines which contain the header information
headers <- read.csv(text=",A,,B,,C,,D,,E,,F,
Field,time,T,time,T,time,T,time,T,time,T,time,T
Acute,0,0,8.3,1,7.5,1,8.6,2,0,0,8.3,4
Ane,9,120,7.7,26,7.9,43,7.8,77,7.9,60,8.2,326
En,15.6,2,12.9,1,0,0,0,0,14.3,1,14.6,4
Fo,9.2,2,0,0,5.4,1,0,0,0,0,7.9,3", header=FALSE, stringsAsFactors=FALSE, nrow=2)

# extract the university names for the 'headers' data.frame
universities <- unlist(headers[1,])
universities <- universities[universities != '']

# create column names from the 'headers' data.frame
vec <- headers[2,][headers[2,] == 'T']
headers[2,][headers[2,] == 'T'] <- paste0(vec, seq_along(vec))
names(DF) <- paste0(headers[2,],headers[1,])

您的数据框现在如下所示：

> DF
   Field timeA  T1 timeB T2 timeC T3 timeD T4 timeE T5 timeF  T6
1: Acute   0.0   0   8.3  1   7.5  1   8.6  2   0.0  0   8.3   4
2:   Ane   9.0 120   7.7 26   7.9 43   7.8 77   7.9 60   8.2 326
3:    En  15.6   2  12.9  1   0.0  0   0.0  0  14.3  1  14.6   4
4:    Fo   9.2   2   0.0  0   5.4  1   0.0  0   0.0  0   7.9   3

因为最好将数据转换成长格式：

library(data.table)
DT <- melt(setDT(DF), id = 1, 
           measure.vars = patterns('^time','^T'),
           variable.name = 'university', 
           value.name = c('time','t')
           )[, university := universities[university]][]

现在您的数据如下所示：

> DT
    Field university time   t
 1: Acute          A  0.0   0
 2:   Ane          A  9.0 120
 3:    En          A 15.6   2
 4:    Fo          A  9.2   2
 5: Acute          B  8.3   1
 6:   Ane          B  7.7  26
 7:    En          B 12.9   1
 8:    Fo          B  0.0   0
 9: Acute          C  7.5   1
10:   Ane          C  7.9  43
11:    En          C  0.0   0
12:    Fo          C  5.4   1
13: Acute          D  8.6   2
14:   Ane          D  7.8  77
15:    En          D  0.0   0
16:    Fo          D  0.0   0
17: Acute          E  0.0   0
18:   Ane          E  7.9  60
19:    En          E 14.3   1
20:    Fo          E  0.0   0
21: Acute          F  8.3   4
22:   Ane          F  8.2 326
23:    En          F 14.6   4
24:    Fo          F  7.9   3

现在您可以选择所需的信息：

 DT[university == 'A' & Field == 'Ane']

给出：

   Field university time   t
1:   Ane          A    9 120

---

几个dplyr 过滤数据的例子：

library(dplyr)
DT %>% 
  filter(Field=="En" & t > 1)

给予：

  Field university time t
1    En          A 15.6 2
2    En          F 14.6 4

或者：

DT %>%
  arrange(desc(time)) %>%
  filter(time < 14 & t > 3)

给予：

  Field university time   t
1   Ane          A  9.0 120
2 Acute          F  8.3   4
3   Ane          F  8.2 326
4   Ane          C  7.9  43
5   Ane          E  7.9  60
6   Ane          D  7.8  77
7   Ane          B  7.7  26

Answer 2

改变你的交叉表

,A,,B,,C,,D,,E,,F,
Field,time,T,time,T,time,T,time,T,time,T,time,T
Acute,0,0,8.3,1,7.5,1,8.6,2,0,0,8.3,4
Ane,9,120,7.7,26,7.9,43,7.8,77,7.9,60,8.2,326
En,15.6,2,12.9,1,0,0,0,0,14.3,1,14.6,4
Fo,9.2,2,0,0,5.4,1,0,0,0,0,7.9,3

转换成直接的数据格式

Field,time,T,Experiment
Acut,0,0,A
An,9,120,A
En,15.6,2,A
Fo,9.2,2,A
Acute,8.3,1,B
An,7.7,26,B
En,12.9,1,B
Fo,0,0,B
Acute,7.5,1,C
An,7.9,43,C
En,0,0,C
Fo,5.4,1,C
Acute,8.6,2,D
An,7.8,77,D
En,0,0,D
Fo,0,0,D
Acute,0,0,E
An,7.9,60,E
En,14.3,1,E
Fo,0,0,E
Acute,8.3,4,F
An,8.2,326,F
En,14.6,4,F
Fo,7.9,3,F

我在哪里使用 Vim.csv 插件和可视块模式。

多种选择方式

在将数据整理成易于格式化的直表（不是交叉表）后，这很容易以多种方式完成，我更喜欢 SQL。我在下面演示了 SQLDDF 包，它在处理大数据时效率非常低，但它很小，所以它可以工作。

除了read.csv 等非常低效的内置函数，我会参考data.table 包中非常有效的fread 来读取文件。

SQLDF

> library(data.table);
> a<-fread("~/DS_straight_table.csv");
> sqldf("select time from a where Experiment='A' and Field='An'")
  time
1    9

其他不带sqldf的

> library(data.table);
> a<-fread("~/DS_straight_table.csv");
> a[Experiment=='A' & Field=='An'] 
Field time   T Experiment
1:    An    9 120          A

Answer 3

使用“Tall”（直桌）格式和库 dplyr。您的数据在每个字段（实验）中只有一个值。

library(dplyr)    

## this is the more general result
df %>% 
  group_by(Field, Experiment) %>%
  top_n(1, wt = -time)


## example function
getTimes<- function(data, field, experiment) {
  data %>% 
    filter(Field == field, Experiment == experiment) %>%
    top_n(1, wt = -time)
}


getTimes(df, 'An', 'A')

#   Field time   T Experiment
# 1    An    9 120          A