Selecting the row subgroup with the max value of a variable [duplicate] - r

Question

This question already has answers here:
Select the row with the maximum value in each group
(19 answers)
Closed 2 years ago.
so I have a data frame:
ID: YearMon: Var: Count:
1 012007 H 1
1 012007 D 2
1 022007 NA
1 032007 H 1
2 012007 H 1
2 022007 Na
2 022007 D 1
2 032007 NA
How would i go about getting just the max value for each unique ID for a certain YearMon? Ideally it would return:
1 012007 D 2
1 022007 NA
1 032007 H 1
2 012007 H 1
2 022007 D 1
2 032007 NA

Answer 1

Using plyr this should be easily achieved. This will filter by ID and YearMon and return the max value along with the ID and YearMon in a data frame.
library(plyr)
ddply( dat1 , .(ID,YearMon) ,function(x) {
Count = max( x$Count )
data.frame( Count=Count , Var=x[x$Count == Count,"Var"] )
})
In order to return all Columns:
df[ is.na( df$Count ) , "Count" ] <- -9999
df2 <- ddply(df, .(ID,YearMon) , function(x){
Count = max( x$Count )
index = which( x$Count == max( x$Count ))
y <- x[ index ,]
data.frame( y )
})
df2[ df2$Count == -9999, "Count" ] <- NA
This will return your indexing values back to NA as well.

Answer 2

Using data.table, if you have a data table called dt, you can first calculate the max of Count by group, and then just keep the rows where Count is equal to the max for that group:
newdt <- dt[, max.count := max(Count), by=.(ID, YearMon)][Count==max.count,.(ID, YearMon, Var, Count)]

Answer 3

library(dplyr)
dt %>%
group_by(ID, YearMon) %>%
slice(Count %>% which.max)

Answer 4

Lets not forget about aggregate!
#####Clean up data. You need to change your grouping variables to factors and data needs to be numeric####
dat1$Var.[dat1$Var.==1]=""
dat1$Count.<-as.numeric(dat1$Count.)
dat1$ID.<-as.factor(dat1$ID.)
dat1$YearMon.<-as.factor(dat1$YearMon.)
dat1<-dat1[,-3] ###Lets get rid of the Var column as you're not using it.
aggregate(. ~ ID.+YearMon.,data = dat1,FUN=max ) #### Use aggregate. Simple and short code
ID. YearMon. Count.
1 1 12007 2
2 2 12007 1
3 2 22007 1
4 1 32007 1

Answer 5

Another data.table option using if/else. We convert the 'data.frame' to 'data.table' (setDT(df1)), grouped by 'ID', 'YearMon', if all of the 'Count' values are 'NA' we return the Subset of Data.table (.SD) or else we get the index of maximum value of 'Count' and subset the data.table (.SD[which.max(Count)]).
library(data.table)
setDT(df1)[, if(all(is.na(Count))) .SD else .SD[which.max(Count)],.(ID, YearMon)]
# ID YearMon Var Count
#1: 1 12007 D 2
#2: 1 22007 NA
#3: 1 32007 H 1
#4: 2 12007 H 1
#5: 2 22007 D 1
#6: 2 32007 NA
Or another option would be to concatenate the index from which.max and the rows which have all 'NA' for 'Count' grouped by the variables, get the row index (.I) and use that to subset the 'data.table'.
setDT(df1)[df1[, .I[c(which.max(Count), all(is.na(Count)))], .(ID, YearMon)]$V1]
# ID YearMon Var Count
#1: 1 12007 D 2
#2: 1 22007 NA
#3: 1 32007 H 1
#4: 2 12007 H 1
#5: 2 22007 D 1
#6: 2 32007 NA
Or we replace the NA by a very small number, use which.max and subset
setDT(df1)[, .SD[which.max(replace(Count, is.na(Count),-Inf ))], .(ID, YearMon)]
data
df1 <- structure(list(ID = c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L),
YearMon = c(12007L,
12007L, 22007L, 32007L, 12007L, 22007L, 22007L, 32007L), Var = c("H",
"D", "", "H", "H", "", "D", ""), Count = c(1L, 2L, NA, 1L, 1L,
NA, 1L, NA)), .Names = c("ID", "YearMon", "Var", "Count"),
class = "data.frame", row.names = c(NA, -8L))