r数据框的顺序和选择

Question

I would appreciate if someone could give me some direction of how to solve a complex ordering of a matrix and selection of the top 2 elements in each subcategory.

code:

index<-1:14
metric<-c(0.037777,0.041143,0.041043,0.042056,0.043701,0.042169,0.042134,
          0.046565,0.044638,0.036653,0.046221,0.04033,0.045385,0.043873)
cat_1<-c("California Munis","California Munis","California Munis","California Munis",
         "California Munis","California Munis","California Munis","Corporate Bonds",
         "Corporate Bonds","Corporate Bonds","Government Bonds","Government Bonds",
         "High Yield Bonds","High Yield Bonds")
cat_2<-c("California Munis","Corporate Bonds","Corporate Bonds","Government Bonds",
         "High Yield Bonds","High Yield Bonds","High Yield Bonds","High Yield Bonds",
         "High Yield Bonds","High Yield Bonds","California Munis","California Munis",
         "Corporate Bonds","Corporate Bonds")

data<-data.frame(cbind(index,metric,cat_1,cat_2))

which produces the below matrix

Ind Metric     Cat_1                Cat_2
1   0.037777    California Munis    California Munis
2   0.041143    California Munis    Corporate Bonds
3   0.041043    California Munis    Corporate Bonds
4   0.042056    California Munis    Government Bonds
5   0.043701    California Munis    High Yield Bonds
6   0.042169    California Munis    High Yield Bonds
7   0.042134    California Munis    High Yield Bonds
8   0.046565    Corporate Bonds     High Yield Bonds
9   0.044638    Corporate Bonds     High Yield Bonds
10  0.036653    Corporate Bonds     High Yield Bonds
11  0.046221    Government Bonds    California Munis
12  0.04033     Government Bonds    California Munis
13  0.045385    High Yield Bonds    Corporate Bonds
14  0.043873    High Yield Bonds    Corporate Bonds

Given the matrix above I would like to order based on the Cat_1, Cat_2 and Metric. i have tried this:

data[order(data[,3],data[,4],data[,2]),]

However Cat_1 and Cat_2 should be indifferent if their entries are the same. As an example, "California Munis"&"Corporate Bonds"="Corporate Bonds"&"California Munis". the outcome I am looking to get should look like the result in the following matrix

Ind Metric      Cat_1               Cat_2               Selection
1   0.037777    California Munis    California Munis    1
2   0.041143    California Munis    Corporate Bonds     1
3   0.041043    California Munis    Corporate Bonds     2
11  0.046221    Government Bonds    California Munis    1
4   0.042056    California Munis    Government Bonds    2
12  0.04033     Government Bonds    California Munis    
5   0.043701    California Munis    High Yield Bonds    1
6   0.042169    California Munis    High Yield Bonds    2
7   0.042134    California Munis    High Yield Bonds    
8   0.046565    Corporate Bonds     High Yield Bonds    1
13  0.045385    High Yield Bonds    Corporate Bonds     2
9   0.044638    Corporate Bonds     High Yield Bonds    
14  0.043873    High Yield Bonds    Corporate Bonds 
10  0.036653    Corporate Bonds     High Yield Bonds    

The last column presents the selection of the top 2 lines per every subcategory that I need to extract.

Any ideas or code would be highly appreciated.

Thanks

Answer 1

Please abandon the use of data.frame(cbind(...)) . It will only cause you grief.

 newdat <- data[ with( data, 
                order( pmax( as.numeric(cat_1), as.numeric(cat_2) ), 
                       pmin( as.numeric(cat_1), as.numeric(cat_2) ) ,
                     - metric) ) , ]
 newdat$selection <- ave(index, 
                         first=pmax( as.numeric(newdat$cat_1), 
                                     as.numeric(newdat$cat_2) ), 
                        second= pmin( as.numeric(newdat$cat_1), 
                                      as.numeric(newdat$cat_2) ) ,
                         FUN=seq)
#-----------------------------------------
> newdat
   index   metric            cat_1            cat_2 selection
1      1 0.037777 California Munis California Munis         1
2      2 0.041143 California Munis  Corporate Bonds         1
3      3 0.041043 California Munis  Corporate Bonds         2
11    11 0.046221 Government Bonds California Munis         1
4      4 0.042056 California Munis Government Bonds         2
12    12 0.040330 Government Bonds California Munis         3
5      5 0.043701 California Munis High Yield Bonds         1
6      6 0.042169 California Munis High Yield Bonds         2
7      7 0.042134 California Munis High Yield Bonds         3
8      8 0.046565  Corporate Bonds High Yield Bonds         1
13    13 0.045385 High Yield Bonds  Corporate Bonds         2
9      9 0.044638  Corporate Bonds High Yield Bonds         3
14    14 0.043873 High Yield Bonds  Corporate Bonds         4
10    10 0.036653  Corporate Bonds High Yield Bonds         5

The requirement for success here is that the levels in the two cat variables are the same. If not, then make them the same with levels(.) <- union(levels(cat1, levels(cat_2))

Answer 2

I expand on my comment

# introduce combined category
cat3 <- sapply(paste(data$cat_1,data$cat_2,sep=" "),function(x){paste(sort(strsplit(x," ")[[1]]), collapse=" ")})
data$cat_3 <- cat3
# order as desired
data1 <- data[order( cat_3 , -metric), ]
# label and select top 2 in each cat
data1$rankByCat <- unlist(sapply(unique(data1$cat_3), function(mycat, mydf)  {return(1:sum(mydf$cat_3==mycat))}, mydf=data1))
data1[data1$rankByCat < 3, !names(data1)%in%c("cat_3")]

Answer 3

@andrei

I have got the sorting part with the following code:

#concacenate the 2 strings
cat_3<-paste(data[,3],data[,4],sep="  ")

#break the string to 2 (creates a list)
temp_split<-strsplit(cat_3,"  ")

#sort by row
sort_split<-sapply(temp_split,sort)

#bind split
out<-cbind(data,t(sort_split))

Is that the best way to write it?

How would I proceed from here to select the top 2 of each category?

Thanks for the help!