17

I'm using the dplyr package in R and have grouped my data by 3 variables (Year, Site, Brood).

I want to get rid of groups made up of less than 3 rows. For example in the following sample I would like to remove the rows for brood '2'. I have a lot of data to do this with so while I could painstakingly do it by hand it would be so helpful to automate it using R. 

Year Site Brood Parents
1996 A    1     1  
1996 A    1     1  
1996 A    1     0  
1996 A    1     0  
1996 A    2     1      
1996 A    2     0  
1996 A    3     1  
1996 A    3     1  
1996 A    3     1  
1996 A    3     0  
1996 A    3     1

I hope this makes sense and thank you very much in advance for your help! I'm new to R and stackoverflow so apologies if the way I've worded this question isn't very good! Let me know if I need to provide any other information.

Keeley Seymour
  • 281
  • 3
  • 10
  • Is dplyr necessary? Or are solutions in base-R or data.table also appropriate? – Heroka Feb 08 '16 at 14:32
  • 3
    @Heroka `data %>% group_by(Year, Site, Brood) %>% filter(n() >= 3)` why wouldn't you use dplyr? ;) – Mullefa Feb 08 '16 at 14:40
  • 1
    @Mullefa because there are other options, and I'm personally more comfortable with data.table and base-R. But I understand that preferences can vary between persons :P – Heroka Feb 08 '16 at 14:45
  • @Heroka I'm sure those other methods are completely appropriate! Like I say I'm a total rookie and I've simply used dplyr more than base-R/data.table :P – Keeley Seymour Feb 08 '16 at 15:10

3 Answers3

35

One way to do it is to use the magic n() function within filter:

library(dplyr)

my_data <- data.frame(Year=1996, Site="A", Brood=c(1,1,2,2,2))

my_data %>% 
  group_by(Year, Site, Brood) %>% 
  filter(n() >= 3)

The n() function gives the number of rows in the current group (or the number of rows total if there is no grouping).

drhagen
  • 8,331
  • 8
  • 53
  • 82
  • This awesome answer helped me, here is the official dplyr documentation for n() if anyone else needs it since I had a bit of a hard time finding it: https://dplyr.tidyverse.org/reference/n.html The main takeaway is that n() can be used with summarise(), mutate() and filter() – Ricky Apr 29 '20 at 23:49
9

Throwing the data.table approach here to join the party:

library(data.table)
setDT(my_data)
my_data[ , if (.N >= 3L) .SD, by = .(Year, Site, Brood)]
MichaelChirico
  • 33,841
  • 14
  • 113
  • 198
3

You can also do this using base R:

temp <- read.csv(paste(folder,"test.csv", sep=""), head=TRUE, sep=",")
matches <- aggregate(Parents ~ Year + Site + Brood, temp, FUN="length")
temp <- merge(temp, matches, by=c("Year","Site","Brood"))
temp <- temp[temp$Parents.y >= 3, c(1,2,3,4)]
pluke
  • 3,832
  • 5
  • 45
  • 68
  • 2
    Or in a very ugly oneliner: `dat[unlist(sapply(split(dat,list(dat$Year,dat$Site,dat$Brood)),function(x){rep(nrow(x),nrow(x))}))>3,]` – Heroka Feb 08 '16 at 15:17