Grab Bag of 'ggplot2' Functions

Extra geoms and scales for 'ggplot2', including geom_cloud(), a Normal density cloud replacement for errorbars; transforms ssqrt_trans and pseudolog10_trans, which are loglike but appropriate for negative data; interp_trans() and warp_trans() which provide scale transforms based on interpolation; and an infix compose operator for scale transforms.


Build Status codecov.io CRAN Downloads Total

A grab bag of ggplot2 extensions and hacks.

-- Steven E. Pav, [email protected]

Installation

This package can be installed from CRAN (not yet), via drat, or from github:

# via CRAN: (not yet)
# install.packages("ggallin")
# via drat:
if (require(drat)) {
    drat:::add("shabbychef")
    install.packages("ggallin")
}
# get snapshot from github (may be buggy)
if (require(devtools)) {
  install_github('shabbychef/ggallin')
}

geom_cloud

This geom acts nearly as a drop-in replacement for geom_errorbar, converting ymin and ymax into 'clouds' of uncertainty with alpha proportional to normal density.

library(ggplot2)
library(ggallin)
library(dplyr)
 
nobs <- 1000
 
set.seed(2134)
mydat <- data.frame(grp=sample(c(0,1),nobs,replace=TRUE),
                    colfac=sample(letters[1:2],nobs,replace=TRUE),
                    rowfac=sample(letters[10 + (1:3)],nobs,replace=TRUE)) %>%
  mutate(x=seq(0,1,length.out=nobs) + 0.33 * grp) %>%
  mutate(y=0.25*rnorm(nobs) + 2*grp) %>%
  mutate(grp=factor(grp)) %>%
  mutate(se=sqrt(x)) %>%
  mutate(ymin=y-se,ymax=y+se)
 
offs <- 2
ph <- mydat %>%
  mutate(y=y+offs,ymin=ymin+offs,ymax=ymax+offs) %>%
  ggplot(aes(x=x,y=y,ymin=ymin,ymax=ymax,color=grp,fill=grp)) + 
  facet_grid(rowfac ~ colfac) +
  scale_y_sqrt() + geom_line() + 
  geom_cloud(aes(fill=grp),steps=15,max_alpha=0.85,color=NA) +   
  labs(title='geom cloud')
print(ph)
plot of chunk geom_cloud

log-like transforms

The square root transform is a good compromise between raw and logarithmic scales, showing detail across different scales without over-emphasizing very small variation. However, it does not work for negative numbers. Thus a signed square root transform is useful. Along similar lines, the pseudo-log transform accepts negative numbers while providing a good view across magnitudes. Some illustrations:

library(ggplot2)
library(ggallin)
library(dplyr)
 
nobs <- 100
 
# this is a silly example, don't blame me
set.seed(1234)
mydat <- data.frame(x=rnorm(nobs),z=rnorm(nobs)) %>%
  mutate(y=sign(z) * exp(x+z-2)) 
ph <- mydat %>%
  ggplot(aes(x=x,y=y)) + 
  geom_line() + 
  scale_y_continuous(trans=ssqrt_trans)
print(ph)
plot of chunk loglike_trans
ph <- mydat %>%
  ggplot(aes(x=x,y=y)) + 
  geom_line() + 
  scale_y_continuous(trans=pseudolog10_trans)
print(ph)
plot of chunk loglike_trans

interpolated transforms

Scale transforms are useful for 'straightening out' crooked data graphically. Sometimes these transforms can not be expressed functionally but instead rely on data. In this case we can imagine that we have some paired data that provide the transformation x -> y. We provide a scale transformation that supports linear interpolation. We also provide another scale transformation that accepts x and positive 'weights' w, and computes y by taking the cumulative sum of weights, called a 'warp' transformation.

Here we illustrate the warp transformation by plotting the cumulative return of the 'UMD' factor against a time scale that is uniform in cumulative daily VIX (whatever that means):

library(ggplot2)
library(ggallin)
library(dplyr)
library(aqfb.data)
library(scales)
 
data(dvix)
data(dff4)
 
rr_to_nav <- function(x) {
  exp(cumsum(log(1 + x)))
}
 
rets <- dff4 %>%
  as.data.frame() %>%
  tibble::rownames_to_column(var='date') %>%
  inner_join(dvix %>%
             as.data.frame() %>%
             setNames(c('VIX')) %>%
             tibble::rownames_to_column(var='date'),by='date') %>%
  mutate(date=as.Date(date,format='%Y-%m-%d')) %>%
  mutate(UMD_nav=rr_to_nav(0.01*UMD),
         SMB_nav=rr_to_nav(0.01*SMB),
         HML_nav=rr_to_nav(0.01*HML))
 
  ph <- rets %>%
    ggplot(aes(x=date,y=UMD_nav)) +
    geom_line() +
    labs(y='UMD cumulative return') + 
    labs(x='regular date scale')
  print(ph)
plot of chunk interp_trans
  # select breaks automagically
  ph <- rets %>%
    ggplot(aes(x=date,y=UMD_nav)) +
    geom_line() + 
    scale_x_continuous(trans=warp_trans(x=rets$date,w=rets$VIX)) +
    labs(y='UMD cumulative return') + 
    labs(x='warped date scale')
  print(ph)
plot of chunk interp_trans
  # force decade breaks:
  ph <- rets %>%
    ggplot(aes(x=date,y=UMD_nav)) +
    geom_line() + 
    scale_x_continuous(trans=warp_trans(x=rets$date,w=rets$VIX,
                                        breaks=scales::date_breaks('10 years'),
                                        format=scales::date_format('%Y'))) +
labs(y='UMD cumulative return') + 
labs(x='warped date scale')
print(ph)
plot of chunk interp_trans
# reverse scale as well (see composition of transforms)
ph <- rets %>%
  ggplot(aes(x=date,y=UMD_nav)) +
  geom_line() + 
  scale_x_continuous(trans=scales::reverse_trans() %of% warp_trans(x=rets$date,w=rets$VIX)) +
  labs(y='UMD cumulative return') + 
  labs(x='reversed, warped date scale')
print(ph)
plot of chunk interp_trans

composition of transforms

The %of% binary operator supports composition of scale transformations. This is most useful for composing reverse scales with other transforms:

library(ggplot2)
library(ggallin)
 
#  reverse and log scale
set.seed(1234)
ph <- ggplot(data.frame(x=rnorm(100),y=exp(rnorm(100,mean=-2,sd=4))),aes(x=x,y=y)) + 
  geom_point() + 
  scale_y_continuous(trans=scales::reverse_trans() %of% scales::log10_trans()) +
  labs(title='reversed and log scaled y')
print(ph)
plot of chunk compose_trans

News

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("ggallin")

0.1.1 by Steven E. Pav, 2 years ago


https://github.com/shabbychef/ggallin


Report a bug at https://github.com/shabbychef/ggallin/issues


Browse source code at https://github.com/cran/ggallin


Authors: Steven E. Pav [aut, cre]


Documentation:   PDF Manual  


LGPL-3 license


Imports scales, grid

Depends on ggplot2

Suggests knitr, testthat


See at CRAN