使用数据帧列表的规则合并数据

Question

我有一个数据框，需要根据定义的组进行折叠。数据由数百个组组成。每个组可以有2-5行的任何地方。为了简单起见，我的例子显示了3组2-4行。

我想扁平每个组内的复制品。对于组中的每一列，我想返回不是NA的最大出现值。问题在于如何在平局的情况下做什么。对于关系，我需要根据绑定的值的类型设置自定义规则。一个潜在的绝望选择是将绑定值粘贴在一起，用逗号分隔，我可以用find/replace方式处理它们。

为了获得最大值，我可以使用max函数。有关如何处理关系的建议？

#Input Data Example 
> data 
    Group Loc1 Loc2 Loc3 Loc4 
1 Group1 A/B A/A B/B NA 
2 Group1 A/B A/A B/B A/A 
3 Group1 A/A A/A A/A NA 
4 Group1 A/A A/A A/A NA 
5 Group2 A/A NA C/C B/B 
6 Group2 B/B A/A C/C B/B 
7 Group2 B/B A/A C/C B/B 
8 Group3 B/B B/B NA B/B 
9 Group3 B/B B/B NA A/A 

#Desired Collapsed Output 
> data.collapsed 
    Group Loc1 Loc2 Loc3 Loc4 
1 Group1 NA A/A A/B A/A 
2 Group2 B/B A/A C/C B/B 
3 Group3 B/B B/B NA A/B 

最终代码（更新一月27,2015）

library(data.table) 
#Data Frame 
#Each group has replicates of data that need to be collapsed to make a consensus data replicate 
data = rbind(c("Group1","A/B", "A/A","B/B",NA), c("Group1","A/B", "A/A","B/B","A/A"), c("Group1","A/A", "A/A","A/A",NA), 
     c("Group1","A/A", "A/A","A/A",NA), c("Group2","A/A", NA,"C/C","B/B"), c("Group2","B/B", "A/A","C/C","B/B"), 
     c("Group2","B/B", "A/A","C/C","B/B"), c("Group3","B/B", "B/B",NA,"B/B"), c("Group3","B/B", "B/B",NA,"A/A")) 
colnames(data) = c("Group", "Loc1", "Loc2", "Loc3", "Loc4") 
data = as.data.frame(data) 
data 

#Define acceptable value types; these could be used to define what to do in the case of a tie 
same.letter = c("A/A","B/B","C/C") 
diff.letter = c("A/B","A/C","B/C") 

#Function for collapsing data with rules 
RepMerge = function(col) { 
    z = table(col); 
    z.max = which(z==max(z)); 

    ifelse(length(z.max) > 2, "NA", #if tied between more than 2 different values, report NA 
     ifelse(length(z.max) == 1, names(z)[z.max], #if one max value, report that value 
      ifelse(length(z.max) == 2 & names(z)[z.max][1] %in% same.letter & names(z)[z.max][2] %in% same.letter, paste(substring(names(z)[z.max][1],1,1),substring(names(z)[z.max][2],1,1), sep="/"), #if both max values are different but are in 'same.letter', report a combination 
       ifelse(length(z.max) == 2 & names(z)[z.max][1] %in% diff.letter | names(z)[z.max][2] %in% diff.letter, "NA", "Check Code")))) #if one of the max values is in diff.letter, report NA. If no cases fit the above, report "Check Code" 
} 

setDT(data)[,lapply(.SD,RepMerge),Group] # run function to collapse the data 

谢谢 SC2

Answer 1

这是基于data.table解决方案：

library(data.table) 
setDT(data)[,lapply(.SD,function(cl) {z<-table(cl);z.max<-which(z==max(z));ifelse(length(z.max)>1,"NA",names(z)[z.max])}),Group] 

# Group Loc1 Loc2 Loc3 Loc4 
#1: Group1 NA A/A NA NA 
#2: Group2 B/B A/A C/C B/B 
#3: Group3 B/B B/B NA NA 

通过修改ifelse ，你可以设置理想的规则来处理关系和来港。 PS：你很好奇max函数为什么会忽略代码中的NAs。这是因为你的数据表包含字符串'NA'，而不是实际的NA s。