Minzhao Liu

notes dplyr

Notes for dplyr

http://cran.r-project.org/web/packages/dplyr/vignettes/introduction.html

Basic verbs

filter()

subset of dataset

filter(dat, month == 1 | month == 2)

arrange()

sort the dataset

arrange(dat, year, month, date)

select()

select columns

select(dat, col1, col2)
select(dat, col1:col5)
select(dat, -(col1:col5))

new feature for ddply 0.1.2:

select(df, starts_with("xyz_"))
select(df, contains("xyz_"))
select(df, ends_with("xyz_"))
select(df, matches("xyz.*"))
select(df, num_range(x1:xN)

mutate()

add new columns

mutate(dat, newvar = col1 - col2, newvar2 = mean(x))

summarise()

collapse data to a single row

summarise(dat, meanx = mean(x))

Grouped operations

bycol1 = group_by(dat, col1)
newsummary = summarise(bycol1,
	count = n(),
	meanx = mean(x))
newsummary = filter(newsummary, count > 20, meanx < 200)

plot:

ggplot(newsummary, aes(x, y)) + geom_point(aes(y),) + geom_smooth() + scale_size_area()

can use aggregate function like min, max, sd, IQR

n()
count_distinct(x)
first(x)
last(x)
nth(x, n)

can group by couples of levels

group_by(dat, year, month, date)

Window functions

ranking and ordering

x = c(1, 1, 2,2,2)

row_number(): 1,2,3,4,5; rank
min_rank(): 1,1, 3,3,3
dense_rank(): 1,1,2,2,2
cume_dist(): 0.4, 0.4, 1, 1, 1; proportion of values less than or equal to the current value
percent_rank(): 0, 0, 0.5, 0.5, 0.5;  percentage of the rank
ntile():  divides the data up into n evenly sized buckets
desc() : to rank from highest to lowest.

offsets

lead(): next; 2,3,4,5,NA
lag(): previous, NA, 1,2,3,4

other three families are variations on familiar aggregate functions

cumsum()
cummin()
cummax()
cumall()
cumany()
cummean()

example:

all record for a player after they have played 150 games

filter(players, cumany(G > 150))

order_by

use another variable to order

right <- mutate(scrambled, running = order_by(year, cumsum(value)))

Published

25 January 2014

Modified

liuminzhao 02/25/2014 16:15:01

Tags