Streamgraphs in R?

Question

How can I implement Streamgraphs in R?

Streamgraphs are a variant of stacked graphs and an improvement on Havre et al.'s ThemeRiver in the way the baseline is chosen, layer ordering, and color choice.

Example:

Reference: http://www.leebyron.com/else/streamgraph/

Answer 1

I wrote a solution using lattice::xyplot. The code is at my spacetimeVis repository.

The next example use this data set:

library(lattice)
library(zoo)
library(colorspace)

nCols <- ncol(unemployUSA)
pal <- rainbow_hcl(nCols, c=70, l=75, start=30, end=300)
myTheme <- custom.theme(fill=pal, lwd=0.2)

xyplot(unemployUSA, superpose=TRUE, auto.key=FALSE,
       panel=panel.flow, prepanel=prepanel.flow,
       origin='themeRiver', scales=list(y=list(draw=FALSE)),
       par.settings=myTheme)

It produces this image.

xyplot needs two functions to work: panel.flow and prepanel.flow:

panel.flow <- function(x, y, groups, origin, ...){
  dat <- data.frame(x=x, y=y, groups=groups)
  nVars <- nlevels(groups)
  groupLevels <- levels(groups)

  ## From long to wide
  yWide <- unstack(dat, y~groups)
  ## Where are the maxima of each variable located? We will use
  ## them to position labels.
  idxMaxes <- apply(yWide, 2, which.max)

  ##Origin calculated following Havr.eHetzler.ea2002
  if (origin=='themeRiver') origin= -1/2*rowSums(yWide)
  else origin=0 
  yWide <- cbind(origin=origin, yWide)
  ## Cumulative sums to define the polygon
  yCumSum <- t(apply(yWide, 1, cumsum))
  Y <- as.data.frame(sapply(seq_len(nVars),
                            function(iCol)c(yCumSum[,iCol+1],
                                            rev(yCumSum[,iCol]))))
  names(Y) <- levels(groups)
  ## Back to long format, since xyplot works that way
  y <- stack(Y)$values

  ## Similar but easier for x
  xWide <- unstack(dat, x~groups)
  x <- rep(c(xWide[,1], rev(xWide[,1])), nVars)
  ## Groups repeated twice (upper and lower limits of the polygon)
  groups <- rep(groups, each=2)

  ## Graphical parameters
  superpose.polygon <- trellis.par.get("superpose.polygon")
  col = superpose.polygon$col
  border = superpose.polygon$border 
  lwd = superpose.polygon$lwd 

  ## Draw polygons
  for (i in seq_len(nVars)){
    xi <- x[groups==groupLevels[i]]
    yi <- y[groups==groupLevels[i]]
    panel.polygon(xi, yi, border=border,
                  lwd=lwd, col=col[i])
  }

  ## Print labels
  for (i in seq_len(nVars)){
    xi <- x[groups==groupLevels[i]]
    yi <- y[groups==groupLevels[i]]
    N <- length(xi)/2
    ## Height available for the label
    h <- unit(yi[idxMaxes[i]], 'native') -
      unit(yi[idxMaxes[i] + 2*(N-idxMaxes[i]) +1], 'native')
    ##...converted to "char" units
    hChar <- convertHeight(h, 'char', TRUE)
    ## If there is enough space and we are not at the first or
    ## last variable, then the label is printed inside the polygon.
    if((hChar >= 1) && !(i %in% c(1, nVars))){
      grid.text(groupLevels[i],
                xi[idxMaxes[i]],
                (yi[idxMaxes[i]] +
                 yi[idxMaxes[i] + 2*(N-idxMaxes[i]) +1])/2,
                gp = gpar(col='white', alpha=0.7, cex=0.7),
                default.units='native')
    } else {
      ## Elsewhere, the label is printed outside

      grid.text(groupLevels[i],
                xi[N],
                (yi[N] + yi[N+1])/2,
                gp=gpar(col=col[i], cex=0.7),
                just='left', default.units='native')
    }
  }
}

prepanel.flow <- function(x, y, groups, origin,...){
  dat <- data.frame(x=x, y=y, groups=groups)
  nVars <- nlevels(groups)
  groupLevels <- levels(groups)
  yWide <- unstack(dat, y~groups)
  if (origin=='themeRiver') origin= -1/2*rowSums(yWide)
  else origin=0
  yWide <- cbind(origin=origin, yWide)
  yCumSum <- t(apply(yWide, 1, cumsum))

  list(xlim=range(x),
       ylim=c(min(yCumSum[,1]), max(yCumSum[,nVars+1])),
       dx=diff(x),
       dy=diff(c(yCumSum[,-1])))
}

Answer 2

Maybe something like this with ggplot2. I'm going to edit it later and will also upload the csv data somewhere sensible.

Couple of issues I need to think about:

1. Getting the y value from the smoothed graph so you can overplot the name for high grossing movies
2. Adding a 'wave' to the x-axis as per your example.

Both should be OK to do with a bit of thought. Sadly the interactivity will be tricky. Maybe will have a look at googleVis.

## PRE-REQS
require(plyr)
require(ggplot2)

## GET SOME BASIC DATA
films<-read.csv("box.csv")

## ALL OF THIS IS FAKING DATA
get_dist<-function(n,g){

  dist<-g-(abs(sort(g-abs(rnorm(n,g,g*runif(1))))))
  dist<-c(0,dist-min(dist),0)
  dist<-dist*g/sum(dist)
  return(dist)
}

get_dates<-function(w){
  start<-as.Date("01-01-00",format="%d-%m-%y")+ceiling(runif(1)*365)
  return(start+w)
}

films$WEEKS<-ceiling(runif(1)*10)+6
f<-ddply(films,.(RANK),function(df)expand.grid(RANK=df$RANK,WEEKGROSS=get_dist(df$WEEKS,df$GROSS)))
weekly<-merge(films,f,by=("RANK"))


## GENERATE THE PLOT DATA
plot.data<-ddply(weekly,.(RANK),summarise,NAME=NAME,WEEKDATE=get_dates(seq_along(WEEKS)*7),WEEKGROSS=ifelse(RANK %% 2 == 0,-WEEKGROSS,WEEKGROSS),GROSS=GROSS)

g<-ggplot() + 

  geom_area(data=plot.data[plot.data$WEEKGROSS>=0,],
            aes(x=WEEKDATE,
                ymin=0,
                y=WEEKGROSS,
                group=NAME,
                fill=cut(GROSS,c(seq(0,1000,100),Inf)))
            ,alpha=0.5,
            stat="smooth", fullrange=T,n=1000,
            colour="white",
            size=0.25,alpha=0.5) +

  geom_area(data=plot.data[plot.data$WEEKGROSS<0,],
            aes(x=WEEKDATE,
                ymin=0,
                y=WEEKGROSS,
                group=NAME,
                fill=cut(GROSS,c(seq(0,1000,100),Inf)))
            ,alpha=0.5,
            stat="smooth", fullrange=T,n=1000,
            colour="white",
            size=0.25,alpha=0.5) +

  theme_bw() + 
  scale_fill_brewer(palette="RdPu",name="Gross\nEUR (M)") +
  ylab("") + xlab("")

b<-ggplot_build(g)$data[[1]]
b.ymax<-max(b$y)


## MAKE LABELS FOR GROSS > 450M
labels<-ddply(plot.data[plot.data$GROSS>450,],.(RANK,NAME),summarise,x=median(WEEKDATE),y=ifelse(sum(WEEKGROSS)>0,b.ymax,-b.ymax),GROSS=max(GROSS))
labels<-ddply(labels,.(y>0),transform,NAME=paste(NAME,GROSS),y=(y*1.1)+((seq_along(y)*20*(y/abs(y)))))

## PLOT
g + 
  geom_segment(data=labels,aes(x=x,xend=x,y=0,yend=y,label=NAME),size=0.5,linetype=2,color="purple",alpha=0.5) +
  geom_text(data=labels,aes(x,y,label=NAME),size=3)

Here's a dput() of the films df if anyone wants to play with it:

structure(list(RANK = 1:50, NAME = structure(c(2L, 45L, 18L, 
                                               33L, 32L, 29L, 34L, 23L, 4L, 21L, 38L, 46L, 15L, 36L, 26L, 49L, 
                                               16L, 8L, 5L, 31L, 17L, 27L, 41L, 3L, 48L, 40L, 28L, 1L, 6L, 24L, 
                                               47L, 13L, 10L, 12L, 39L, 14L, 30L, 20L, 22L, 11L, 19L, 25L, 35L, 
                                               9L, 43L, 44L, 37L, 7L, 42L, 50L), .Label = c("Alice in Wonderland", 
                                                                                            "Avatar", "Despicable Me 2", "E.T.", "Finding Nemo", "Forrest Gump", 
                                                                                            "Harry Potter and the Deathly Hallows Part 1", "Harry Potter and the Deathly Hallows Part 2", 
                                                                                            "Harry Potter and the Half-Blood Prince", "Harry Potter and the Sorcerer's Stone", 
                                                                                            "Independence Day", "Indiana Jones and the Kingdom of the Crystal Skull", 
                                                                                            "Iron Man", "Iron Man 2", "Iron Man 3", "Jurassic Park", "LOTR: The Return of the King", 
                                                                                            "Marvel's The Avengers", "Pirates of the Caribbean", "Pirates of the Caribbean: At World's End", 
                                                                                            "Pirates of the Caribbean: Dead Man's Chest", "Return of the Jedi", 
                                                                                            "Shrek 2", "Shrek the Third", "Skyfall", "Spider-Man", "Spider-Man 2", 
                                                                                            "Spider-Man 3", "Star Wars", "Star Wars: Episode II -- Attack of the Clones", 
                                                                                            "Star Wars: Episode III", "Star Wars: The Phantom Menace", "The Dark Knight", 
                                                                                            "The Dark Knight Rises", "The Hobbit: An Unexpected Journey", 
                                                                                            "The Hunger Games", "The Hunger Games: Catching Fire", "The Lion King", 
                                                                                            "The Lord of the Rings: The Fellowship of the Ring", "The Lord of the Rings: The Two Towers", 
                                                                                            "The Passion of the Christ", "The Sixth Sense", "The Twilight Saga: Eclipse", 
                                                                                            "The Twilight Saga: New Moon", "Titanic", "Toy Story 3", "Transformers", 
                                                                                            "Transformers: Dark of the Moon", "Transformers: Revenge of the Fallen", 
                                                                                            "Up"), class = "factor"), YEAR = c(2009L, 1997L, 2012L, 2008L, 
                                                                                                                               1999L, 1977L, 2012L, 2004L, 1982L, 2006L, 1994L, 2010L, 2013L, 
                                                                                                                               2012L, 2002L, 2009L, 1993L, 2011L, 2003L, 2005L, 2003L, 2004L, 
                                                                                                                               2004L, 2013L, 2011L, 2002L, 2007L, 2010L, 1994L, 2007L, 2007L, 
                                                                                                                               2008L, 2001L, 2008L, 2001L, 2010L, 2002L, 2007L, 1983L, 1996L, 
                                                                                                                               2003L, 2012L, 2012L, 2009L, 2010L, 2009L, 2013L, 2010L, 1999L, 
                                                                                                                               2009L), GROSS = c(760.5, 658.6, 623.4, 533.3, 474.5, 460.9, 448.1, 
                                                                                                                                                 436.5, 434.9, 423.3, 422.7, 415, 409, 408, 403.7, 402.1, 395.8, 
                                                                                                                                                 381, 380.8, 380.2, 377, 373.4, 370.3, 366.9, 352.4, 340.5, 336.5, 
                                                                                                                                                 334.2, 329.7, 321, 319.1, 318.3, 317.6, 317, 313.8, 312.1, 310.7, 
                                                                                                                                                 309.4, 309.1, 306.1, 305.4, 304.4, 303, 301.9, 300.5, 296.6, 
                                                                                                                                                 296.3, 295, 293.5, 293), WEEKS = c(9, 9, 9, 9, 9, 9, 9, 9, 9, 
                                                                                                                                                                                    9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 
                                                                                                                                                                                    9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9)), .Names = c("RANK", 
                                                                                                                                                                                                                                                             "NAME", "YEAR", "GROSS", "WEEKS"), row.names = c(NA, -50L), class = "data.frame")

Answer 3

I wrote a function plot.stacked a while back that might be able to help you out.

The function is:

plot.stacked <- function(x,y, ylab="", xlab="", ncol=1, xlim=range(x, na.rm=T), ylim=c(0, 1.2*max(rowSums(y), na.rm=T)), border = NULL, col=rainbow(length(y[1,]))){

    plot(x,y[,1], ylab=ylab, xlab=xlab, ylim=ylim, xaxs="i", yaxs="i", xlim=xlim, t="n")
    bottom=0*y[,1]
    for(i in 1:length(y[1,])){
        top=rowSums(as.matrix(y[,1:i]))
        polygon(c(x, rev(x)), c(top, rev(bottom)), border=border, col=col[i])
        bottom=top
    }
    abline(h=seq(0,200000, 10000), lty=3, col="grey")
    legend("topleft", rev(colnames(y)), ncol=ncol, inset = 0, fill=rev(col), bty="0", bg="white", cex=0.8, col=col)
    box()
}

Here's an example data set and a plot:

set.seed(1)
m <- 500
n <- 15
x <- seq(m)
y <- matrix(0, nrow=m, ncol=n)
colnames(y) <- seq(n)
for(i in seq(ncol(y))){
    mu <- runif(1, min=0.25*m, max=0.75*m)
    SD <- runif(1, min=5, max=30)
    TMP <- rnorm(1000, mean=mu, sd=SD)
    HIST <- hist(TMP, breaks=c(0,x), plot=FALSE)
    fit <- smooth.spline(HIST$counts ~ HIST$mids)
    y[,i] <- fit$y
}

    plot.stacked(x,y)

I can imagine that you would just need to adjust the definition of the polygon "bottom" to get the plot you desire.

Update:

I've had another go at making the stream plot and believe I have more or less reproduced the idea in the function plot.stream, available in this gist and also copied in at the bottom of this post. At this link I show more detail of its use, but here's a basic example:

library(devtools)
source_url('https://gist.github.com/menugget/7864454/raw/f698da873766347d837865eecfa726cdf52a6c40/plot.stream.4.R')

set.seed(1)
m <- 500
n <- 50
x <- seq(m)
y <- matrix(0, nrow=m, ncol=n)
colnames(y) <- seq(n)
for(i in seq(ncol(y))){
    mu <- runif(1, min=0.25*m, max=0.75*m)
    SD <- runif(1, min=5, max=30)
    TMP <- rnorm(1000, mean=mu, sd=SD)
    HIST <- hist(TMP, breaks=c(0,x), plot=FALSE)
    fit <- smooth.spline(HIST$counts ~ HIST$mids)
    y[,i] <- fit$y
}
y <- replace(y, y<0.01, 0)

#order by when 1st value occurs
ord <- order(apply(y, 2, function(r) min(which(r>0))))
y2 <- y[, ord]
COLS <- rainbow(ncol(y2))

png("stream.png", res=400, units="in", width=12, height=4)
par(mar=c(0,0,0,0), bty="n")
plot.stream(x,y2, axes=FALSE, xlim=c(100, 400), xaxs="i", center=TRUE, spar=0.2, frac.rand=0.1, col=COLS, border=1, lwd=0.1)
dev.off()

Code for plot.stream()

#plot.stream makes a "stream plot" where each y series is plotted 
#as stacked filled polygons on alternating sides of a baseline.
#
#Arguments include:
#'x' - a vector of values
#'y' - a matrix of data series (columns) corresponding to x
#'order.method' = c("as.is", "max", "first") 
#  "as.is" - plot in order of y column
#  "max" - plot in order of when each y series reaches maximum value
#  "first" - plot in order of when each y series first value > 0
#'center' - if TRUE, the stacked polygons will be centered so that the middle,
#i.e. baseline ("g0"), of the stream is approximately equal to zero. 
#Centering is done before the addition of random wiggle to the baseline. 
#'frac.rand' - fraction of the overall data "stream" range used to define the range of
#random wiggle (uniform distrubution) to be added to the baseline 'g0'
#'spar' - setting for smooth.spline function to make a smoothed version of baseline "g0"
#'col' - fill colors for polygons corresponding to y columns (will recycle)
#'border' - border colors for polygons corresponding to y columns (will recycle) (see ?polygon for details)
#'lwd' - border line width for polygons corresponding to y columns (will recycle)
#'...' - other plot arguments
plot.stream <- function(
    x, y, 
    order.method = "as.is", frac.rand=0.1, spar=0.2,
    center=TRUE,
    ylab="", xlab="",  
    border = NULL, lwd=1, 
    col=rainbow(length(y[1,])),
    ylim=NULL, 
    ...
){

if(sum(y < 0) > 0) error("y cannot contain negative numbers")

if(is.null(border)) border <- par("fg")
border <- as.vector(matrix(border, nrow=ncol(y), ncol=1))
col <- as.vector(matrix(col, nrow=ncol(y), ncol=1))
lwd <- as.vector(matrix(lwd, nrow=ncol(y), ncol=1))

if(order.method == "max") {
    ord <- order(apply(y, 2, which.max))
    y <- y[, ord]
    col <- col[ord]
    border <- border[ord]
}

if(order.method == "first") {
    ord <- order(apply(y, 2, function(x) min(which(r>0))))
    y <- y[, ord]
    col <- col[ord]
    border <- border[ord]
}

bottom.old <- rep(0, length(x))
top.old <- rep(0, length(x))
polys <- vector(mode="list", ncol(y))
for(i in seq(polys)){
    if(i %% 2 == 1){ #if odd
        top.new <- top.old + y[,i]
        polys[[i]] <- list(x=c(x, rev(x)), y=c(top.old, rev(top.new)))
        top.old <- top.new
    }
    if(i %% 2 == 0){ #if even
        bottom.new <- bottom.old - y[,i]
        polys[[i]] <- list(x=c(x, rev(x)), y=c(bottom.old, rev(bottom.new)))
        bottom.old <- bottom.new
    }
}

ylim.tmp <- range(sapply(polys, function(x) range(x$y, na.rm=TRUE)), na.rm=TRUE)
outer.lims <- sapply(polys, function(r) rev(r$y[(length(r$y)/2+1):length(r$y)]))
mid <- apply(outer.lims, 1, function(r) mean(c(max(r, na.rm=TRUE), min(r, na.rm=TRUE)), na.rm=TRUE))

#center and wiggle
if(center) {
    g0 <- -mid + runif(length(x), min=frac.rand*ylim.tmp[1], max=frac.rand*ylim.tmp[2])
} else {
    g0 <- runif(length(x), min=frac.rand*ylim.tmp[1], max=frac.rand*ylim.tmp[2])
}

fit <- smooth.spline(g0 ~ x, spar=spar)

for(i in seq(polys)){
    polys[[i]]$y <- polys[[i]]$y + c(fitted(fit), rev(fitted(fit)))
}

if(is.null(ylim)) ylim <- range(sapply(polys, function(x) range(x$y, na.rm=TRUE)), na.rm=TRUE)
plot(x,y[,1], ylab=ylab, xlab=xlab, ylim=ylim, t="n", ...)
for(i in seq(polys)){
    polygon(polys[[i]], border=border[i], col=col[i], lwd=lwd[i])
}

}

Answer 4

Adding one line to Marc in the box's nifty code will get you much closer. (Getting the rest of the way will just be a matter of setting fill colors based on the max height of each curve.)

## reorder the columns so each curve first appears behind previous curves
## when it first becomes the tallest curve on the landscape
y <- y[, unique(apply(y, 1, which.max))]

## Use plot.stacked() from Marc's post
plot.stacked(x,y)

Answer 5

These days there's a streamgraphs htmlwidget:

https://hrbrmstr.github.io/streamgraph/

devtools::install_github("hrbrmstr/streamgraph")

library(streamgraph)

streamgraph(data, key, value, date, width = NULL, height = NULL,
  offset = "silhouette", interpolate = "cardinal", interactive = TRUE,
  scale = "date", top = 20, right = 40, bottom = 30, left = 50)

It produces really pretty charts and is even interactive.

Edit

Another option is to use ggTimeSeries which uses the syntax of ggplot2:

# creating some data
library(ggTimeSeries)
library(ggplot2)

set.seed(10)
dfData = data.frame(
   Time = 1:1000,
   Signal = abs(
      c(
         cumsum(rnorm(1000, 0, 3)), 
         cumsum(rnorm(1000, 0, 4)), 
         cumsum(rnorm(1000, 0, 1)),
         cumsum(rnorm(1000, 0, 2))
      )
   ),
   VariableLabel = c(rep('Class A', 1000),
                     rep('Class B', 1000),
                     rep('Class C', 1000),
                     rep('Class D', 1000))
)

# base plot
ggplot(dfData,
       aes(x = Time,
           y = Signal,
           group = VariableLabel,
           fill = VariableLabel)) +
   stat_steamgraph() +
   theme_bw()