CODEWITHSUNDEEP
rplotggplot2
Patthebug
Patthebug

Reputation: 4787

Stacked area plot not showing negative values properly

I am trying to create a stacked area plot. I have been able to do so for the most part except that one of the variables has negative values and it is not being displayed properly. Here's what my data looks like:

Time    A   B   C   D   E   F   G   H   I   J   K
11/1/2014   53438.819   0   1902.19 0   3620.333    12861.2876  0   315.61  0   34739.4 0
12/1/2014   53626.763   0   1908.88 0   3633.066    12906.5207  0   316.72  0   34861.58    0
1/1/2015    50744.951   0   1806.3  0   3437.831    12212.946   0   299.7   0   32988.17    0
2/1/2015    50807.599   0   1808.53 0   3442.075    12228.0237  0   300.07  0   33028.9 0
3/1/2015    50932.895   0   1812.99 0   3450.564    12258.1792  0   300.81  0   33110.35    0
4/1/2015    7264.046    8489.086    258.5685    1465.2215   25465.54    1748.2606   259.5347    42.9015 243.1085    -20251.22   6521.221
5/1/2015    7226.457    8445.158    257.2305    1457.6395   25203.163   1739.214    258.1918    42.6795 241.8505    -20015.83   6487.476
6/1/2015    7245.251    8467.122    257.8995    1461.4305   24940.787   1743.7373   258.8632    42.7905 242.4795    -19739.96   6504.349
7/1/2015    6906.952    8071.77 245.8575    1393.1925   24678.41    1662.3177   246.7763    40.7925 231.1575    -19720.43   6200.644
8/1/2015    7009.4  8191.496    243.1815    1378.0285   24416.033   1693.5511   244.0902    40.3485 228.6415    -19383.71   6340.736
9/1/2015    7019.042    8202.763    243.516 1379.924    24153.657   1695.8806   244.426 40.404  228.956 -19114.42   6349.457

I am using the following code (after melting the data frame to) the generate the plot:

p <- ggplot(temp, aes( Time, value)) + theme_bw() +ylab('Monthly Revenue') + xlab('') +
  scale_x_date(breaks=x_breaks, labels=x_labels)
p <- p + geom_area(aes(colour = variable, fill= variable), position = 'stack',alpha=0.6) +
  theme(axis.text.y=element_text(hjust=0, angle=0), 
        axis.text.x = element_text(hjust=1, angle=45),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        panel.grid.major.x=element_line(color='grey90',linetype='dashed'),
        panel.grid.major.y=element_line(color='grey90',linetype='dashed'),
        plot.title=element_text(size=20),
        axis.text=element_text(size=15),
        legend.text=element_text(size=15),
        legend.key=element_blank(),
        legend.title=element_blank()) +
  scale_y_continuous(label=thousand_formatter) +
  ggtitle('Any Title') + ylab("Dollars")

and here's the plot that I get:

enter image description here

We can see that the negative values are not being displayed properly. How can we display them properly so that the negative values are clearly distinguishable and are under the X axis?

Any help in this regard would be much appreciated.

EDIT:

I used the following code now (which is just a hack as stacked area plot shows the cumulative sum, this is also the reason I added negative values to the variable B so that the effect of negative values may be offset):

temp <- temp[,c(1,11,2:10,12)]
temp[6:11,3] <- temp[6:11,3] + (-1 * temp[6:11,2])
temp <- melt(temp,id='Time')
p <- ggplot(temp, aes( Time, value)) + theme_bw() +ylab('Monthly Revenue') + xlab('') +
  scale_x_date(breaks=x_breaks, labels=x_labels)
p <- p + geom_area(aes(colour = variable, fill= variable), position = 'stack',alpha=0.6) +
  theme(axis.text.y=element_text(hjust=0, angle=0), 
        axis.text.x = element_text(hjust=1, angle=45),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        panel.grid.major.x=element_line(color='grey90',linetype='dashed'),
        panel.grid.major.y=element_line(color='grey90',linetype='dashed'),
        plot.title=element_text(size=20),
        axis.text=element_text(size=15),
        legend.text=element_text(size=15),
        legend.key=element_blank(),
        legend.title=element_blank()) +
  scale_y_continuous(label=thousand_formatter) +
  ggtitle('Any Title') + ylab("Dollars")
p

and I got the following result:

enter image description here

We can still see that the colors are sorta screwed. Don't know why does pink changes to brown as soon as the values turn negative.

Upvotes: 2

Views: 2121

Answers (2)

MrFlick
MrFlick

Reputation: 206207

I still don't quite see this as a stacked area plot, but you could create two groups for each variable to partitiion the negative and positive values. Here's a very rough sketch of how you might .

Here I assume your sample data is in a data.frame named temp

tt<-melt(temp, id.vars="Time")
tt$pos <- tt$value>=0
gg<-expand.grid(Time=unique(tt$Time), pos=unique(tt$pos), variable=unique(tt$variable))
mm<-merge(tt, gg, all=T)
mm$value[is.na(mm$value)]<-0
mm$grp = interaction(mm$variable, factor(mm$pos, levels=c(TRUE,FALSE)))
mm$ymin<-with(mm, ave(value, pos, Time, FUN=function(x) cumsum(c(0,x[-length(x)]))))
mm$ymax<-with(mm, ave(value, pos, Time, FUN=cumsum))


ggplot(mm, aes(x=Time, fill=variable)) + theme_bw() + 
    geom_ribbon(aes(group=grp, color=variable, ymin=ymin, ymax=ymax), alpha=.6)

basically we just create a ribbon chart and do the stacking ourselves. This will eliminate overlapping.

enter image description here

Upvotes: 2

Jaap
Jaap

Reputation: 83215

Why not use a line plot? It's much better in my opinion, especially with negative values:

library(dplyr)
library(tidyr)
library(ggplot2)

temp <- df %>% gather(type, value, -Time)

ggplot(temp, aes(Time, value, group=type, colour=type)) + 
  geom_line(size=1) +
  theme_bw()

which gives: enter image description here

As you want to make an area plot, I gues you want to show the total as well. You can add that to the plot with:

df$all <- rowSums(df[,-1])

After that you can make a line plot with extra thick line for the total:

ggplot(temp[temp$type!="all",], aes(Time, value, group=type, colour=type)) + 
  geom_line(size=1) +
  geom_line(data=temp[temp$type=="all",], aes(Time, value), colour="black", size=1.5) +
  theme_bw()

which gives: enter image description here

EDIT:

I found a way to hack it into an area plot. Supposing your dataframe before melting is called df, you should change the order of the columns first with:

df <- df[,c(1,11,6,2,3,4,5,7,8,9,10,12)]

than melt it with (supposing you're using tidyr):

temp <- df %>% gather(type, value, -Time)

after that you can create your plot with:

ggplot(temp, aes(Time, value, group=type, colour=type)) + 
  geom_area(aes(colour=type, fill=type), alpha=0.4) +
  theme_bw()

which gives: enter image description here

Upvotes: 2

Related Questions