Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

return multiple columns from data.table aggregation [duplicate]

I would like to use data.table as an alternative to aggregate() or ddply(), as these two methods aren't scaling to large objects as efficiently as hoped. Unfortunately, I haven't figured out how to get vector-returning aggregate functions to generate multiple columns in the result from data.table. For example:

# required packages
library(plyr)
library(data.table)

# simulated data
x <- data.table(value=rnorm(100), g=rep(letters[1:5], each=20))

# ddply output that I would like to get from data.table
ddply(data.frame(x), 'g', function(i) quantile(i$value))

 g        0%        25%          50%       75%     100%
 1 a -1.547495 -0.7842795  0.202456288 0.6098762 2.223530
 2 b -1.366937 -0.4418388 -0.085876995 0.7826863 2.236469
 3 c -2.064510 -0.6411390 -0.257526983 0.3213343 1.039053
 4 d -1.773933 -0.5493362 -0.007549273 0.4835467 2.116601
 5 e -0.780976 -0.2315245  0.194869630 0.6698881 2.207800

# not quite what I am looking for:
x[, quantile(value), by=g]

g           V1
1: a -1.547495345
2: a -0.784279536
3: a  0.202456288
4: a  0.609876241
5: a  2.223529739
6: b -1.366937074
7: b -0.441838791
8: b -0.085876995
9: b  0.782686277
10: b  2.236468703

Essentially, the output from ddply and aggregate are in wide-format, while the output from the data.table is in long format. Is the answer reshaping the data, or some additional arguments to my data.table object?

like image 419
Dylan Avatar asked Aug 31 '25 17:08

Dylan


1 Answers

Try coercing to a list:

> x[, as.list(quantile(value)), by=g]
   g         0%          25%         50%       75%     100%
1: a -1.7507334 -0.632331909  0.07435249 0.7459778 1.428552
2: b -2.2043481 -0.005652353  0.10534325 0.5769475 1.241754
3: c -1.9313985 -1.120737610 -0.26116926 0.6953009 1.360017
4: d -0.7434664 -0.055232431  0.22062823 1.1864389 3.021124
5: e -2.0101657 -0.468674094  0.20209610 0.6286448 2.433152
like image 100
GSee Avatar answered Sep 02 '25 08:09

GSee