opts_chunk$set(warning = FALSE, error = FALSE, message = FALSE, tidy = FALSE)
Dummy datasets (data.frame, and list of data.frames:
df <- iris
ld <- replicate(3, data.frame(x=rnorm(10),y=rnorm(10)),simplify=FALSE)
names(ld) <- LETTERS[1:3]
str(df)
## 'data.frame': 150 obs. of 5 variables:
## $ Sepal.Length: num 5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
## $ Sepal.Width : num 3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
## $ Petal.Length: num 1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
## $ Petal.Width : num 0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
## $ Species : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...
str(ld)
## List of 3
## $ A:'data.frame': 10 obs. of 2 variables:
## ..$ x: num [1:10] 0.318 -0.107 -0.772 -0.899 -0.256 ...
## ..$ y: num [1:10] -1.653 0.749 -0.799 1.283 -1.044 ...
## $ B:'data.frame': 10 obs. of 2 variables:
## ..$ x: num [1:10] -0.292 -0.45 -1.265 -0.642 0.446 ...
## ..$ y: num [1:10] 1.207 0.14 0.969 -0.231 0.376 ...
## $ C:'data.frame': 10 obs. of 2 variables:
## ..$ x: num [1:10] -1.314 -1.892 -1.6 -0.573 0.854 ...
## ..$ y: num [1:10] 1.0367 0.8039 -0.0995 -1.48 -1.2969 ...
Naming convention: **ply functions (namely llply, ldply, laply, ddply, dlply, daply, alply, adply, aaply) follow a convenient nomenclature whereby the first letter refers to the type of input variable (a for array, d for data.frame, l for list) and the second letter the output type. Thus, llply works on each element of a list, and returns a list, while adply works on a slice of an array, and returns a data.frame. Of course the function to apply to each subset should have compatible behaviour.
Here's an example:
llply(ld, transform, z = y^2, .progress = "text")
##
|
| | 0%
|
|====================== | 33%
|
|=========================================== | 67%
|
|=================================================================| 100%
## $A
## x y z
## 1 0.31791 -1.6533 2.7332
## 2 -0.10695 0.7486 0.5603
## 3 -0.77196 -0.7995 0.6392
## 4 -0.89935 1.2832 1.6467
## 5 -0.25630 -1.0443 1.0905
## 6 0.11536 0.8378 0.7019
## 7 0.02464 -0.9568 0.9154
## 8 0.24214 -0.4570 0.2088
## 9 -1.18019 2.0997 4.4086
## 10 0.18165 -1.3804 1.9056
##
## $B
## x y z
## 1 -0.2923 1.20718 1.4572902
## 2 -0.4495 0.13960 0.0194879
## 3 -1.2654 0.96869 0.9383669
## 4 -0.6421 -0.23099 0.0533545
## 5 0.4462 0.37605 0.1414125
## 6 -0.1208 1.70897 2.9205749
## 7 1.5694 -0.65589 0.4301878
## 8 -0.9846 -1.17199 1.3735510
## 9 -0.8279 -0.02682 0.0007191
## 10 -1.0417 -0.46854 0.2195334
##
## $C
## x y z
## 1 -1.3138 1.03667 1.074695
## 2 -1.8917 0.80394 0.646313
## 3 -1.6004 -0.09949 0.009899
## 4 -0.5732 -1.48004 2.190524
## 5 0.8537 -1.29694 1.682041
## 6 0.5151 -0.77693 0.603619
## 7 -1.2700 1.64763 2.714681
## 8 0.2653 -0.13564 0.018398
## 9 0.3490 0.36592 0.133900
## 10 1.0267 1.28048 1.639630
We applied the function transform to each data.frame in the list, adding a new variable z defined as z=y^2. If the calculation is slow, it is sometimes nice to have a progress bar (default is none).
Sometimes the function only has side-effects and we don't care about saving results; a typical example is plotting. There's a special character for this, _
l_ply(ld, plot)
Working on data.frames, one needs to specify how to form the subsets,
ddply(df, c("Species"), summarise, maxsepallength = max(Sepal.Length), petalsd=sd(Petal.Width))
## Species maxsepallength petalsd
## 1 setosa 5.8 0.1054
## 2 versicolor 7.0 0.1978
## 3 virginica 7.9 0.2747
where summarise is a convenient function to create a new data.frame with the arguments (…) that follow in the call.
Similarly, working on arrays, one specifies which margin to slice.
Finally, my favorite is m*ply, which takes multiple input arguments:
peak = function(nu0=100, gamma=10, I0 =1,
fun=c("gauss","lorentz"), offset=0, ...){
fun <- match.arg(fun)
x <- seq(nu0-5*gamma,nu0+5*gamma, length=200)
result <- switch(fun,
gauss = I0*dnorm(x, nu0, gamma, ...),
lorentz = I0*gamma / pi/((x-nu0)^2 + gamma^2))
data.frame(x = x, value = result + offset)
}
# make up some combinations of parameters
params <- expand.grid(nu0=c(100, 120, 130),
gamma = seq(10, 30, length=3),
fun=c("gauss","lorentz"), stringsAsFactors=FALSE)
head(params)
## nu0 gamma fun
## 1 100 10 gauss
## 2 120 10 gauss
## 3 130 10 gauss
## 4 100 20 gauss
## 5 120 20 gauss
## 6 130 20 gauss
# apply to all combinations and combined into a data.frame
all <- mdply(params, peak, offset=2)
str(all)
## 'data.frame': 3600 obs. of 5 variables:
## $ nu0 : num 100 100 100 100 100 100 100 100 100 100 ...
## $ gamma: num 10 10 10 10 10 10 10 10 10 10 ...
## $ fun : chr "gauss" "gauss" "gauss" "gauss" ...
## $ x : num 50 50.5 51 51.5 52 ...
## $ value: num 2 2 2 2 2 ...
The results are in long format, thus easily visualised with ggplot2
library(ggplot2)
ggplot(all, aes(x, value, colour=factor(gamma), linetype=factor(fun))) +
facet_wrap(~nu0, ncol=1, scales="free") +
geom_line() +
theme_minimal() +
labs(colour="Width", linetype="Function")
