Deleting rows that are duplicated in one column based on the conditions of another column

问题

Here is an example of my data set;

Date      Time(GMT)Depth Temp  Salinity Density Phosphate
24/06/2002  1000    1           33.855          0.01
24/06/2002  1000    45          33.827          0.01
01/07/2002  1000    10  13.26   33.104  24.873  0.06
01/07/2002  1000    30  12.01   33.787  25.646  0.13
08/07/2002  1000    5   13.34   33.609  25.248  0.01
08/07/2002  1000    40  12.01   34.258  26.011  1.33
15/07/2002  1000    30  12.04   34.507  26.199  0.01
22/07/2002  1000    5   13.93   33.792  25.269  0.01
22/07/2002  1000    30  11.9    34.438  26.172  0.08
29/07/2002  1000    5   13.23   34.09   25.642  0.01

I want to delete duplicate rows so that I only have one row per date, I want to do this based on the Depth, I would like to keep the row with the greatest (deepest) depth. Any ideas?

回答1:

Lets say you have data in df

df = df[order(df[,'Date'],-df[,'Depth']),]
df = df[!duplicated(df$Date),]

回答2:

Introducing a data.table solution which will be the fastest way to solve this (assuming data is your data set)

library(data.table)
unique(setDT(data)[order(Date, -Depth)], by = "Date")

Just another way:

setDT(data)[data[, .I[which.max(Depth)], by=Date]$V1]

回答3:

This might be not the fastest approach if your data frame is large, but a fairly strightforward one. This might change the order of your data frame and you might need to reorder by e.g. date afterwards. Instead of deleting we split the data by date, in each chunk pick a row with the maximum date and finally join the result back into a data frame

data = split(data, data$Date)
data = lapply(data, function(x) x[which.max(x$Depth), , drop=FALSE])
data = do.call("rbind", data)

回答4:

# First find the maxvalues
maxvals = aggregate(df$Depth~df$Date, FUN=max)
#Now use apply to find the matching rows and separate them out
out = df[apply(maxvals,1,FUN=function(x) which(paste(df$Date,df$Depth) == paste(x[1],x[2]))),]

Does that work for you?

回答5:

You might also use dplyr's arrange() instead of order (I find it more intuitive):

df <- arrange(df, Date, -Depth)
df <- df[!duplicated(df$Date),]

来源：https://stackoverflow.com/questions/24011246/deleting-rows-that-are-duplicated-in-one-column-based-on-the-conditions-of-anoth

标签

date

duplicate-removal