Quick Subset Using data.table

if (!require(data.table)) install.packages("data.table")

## Loading required package: data.table

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:data.table':
## 
##     between, first, last

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(tidyr)
library(data.table)
data(population, package = "tidyr")

I use population data from tidyr package as an example.

glimpse(population)

## Observations: 4,060
## Variables: 3
## $ country    <chr> "Afghanistan", "Afghanistan", "Afghanistan", "Afgha...
## $ year       <int> 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 200...
## $ population <int> 17586073, 18415307, 19021226, 19496836, 19987071, 2...

summary(population)

##    country               year        population       
##  Length:4060        Min.   :1995   Min.   :1.129e+03  
##  Class :character   1st Qu.:1999   1st Qu.:6.029e+05  
##  Mode  :character   Median :2004   Median :5.319e+06  
##                     Mean   :2004   Mean   :3.003e+07  
##                     3rd Qu.:2009   3rd Qu.:1.855e+07  
##                     Max.   :2013   Max.   :1.386e+09

class(population)

## [1] "tbl_df"     "tbl"        "data.frame"

populationDT <- as.data.table(population)
class(populationDT)

## [1] "data.table" "data.frame"

t <- proc.time()
tmp1 <- populationDT[
  , .SD[population < median(population)], by = c("country")
]
proc.time() - t

##    user  system elapsed 
##   0.063   0.000   0.061

t <- proc.time()
tmp2 <- populationDT[
  populationDT[, .I[population < median(population)], by = c("country")]$V1
]
proc.time() - t

##    user  system elapsed 
##   0.009   0.000   0.008

t <- proc.time()
tmp3 <- population %>% 
  group_by(country) %>% 
  filter(population < median(population)) %>% 
  ungroup()
proc.time() - t

##    user  system elapsed 
##   0.053   0.000   0.053

identical(tmp1, tmp2)

## [1] TRUE

setdiff(tmp2, tmp3)

## Empty data.table (0 rows) of 3 cols: country,year,population