R ggplot geom_bar Error: Discrete value supplied to continuous scale

Question

I have proportion response data for 4 experimental groups, with 2 different statistics computed for each group. I want the following figure (which I can achieve):

I obtain this figure with the following code:

Group<-c('a','b','c','d','a','b','c','d')
Statistic<-c('Mean','Mean','Mean','Mean','d','d','d','d')
Val<-c(.75,.83,.79,.69,.5,.02,.1,.3)
dfm2<-data.frame(cbind(Group,Statistic,Val)) 
ggplot(dfm2,aes(x = Group,y = Val)) +    
 geom_bar(aes(fill = Statistic),position = dodge',stat='identity')

However, when I change the limits of the y-axis (to [0,1] since I have proportions) by adding the line of code:

+ scale_y_continuous(limits=c(0, 1))

I get

Error: Discrete value supplied to continuous scale

So I understand this means I have a non-continuous variable. I have tried converting my Statistic variable by using as.numeric() and countless other options but to no avail. If anyone can assist me in solving this problem and/or explain what causes it, I would be most appreciative.

Tenet

Answer 1

The problem is the unnecessary use of cbind inside data.frame. cbind creates a matrix. A matrix must have all values of the same mode (numeric, character, etc.). Since at least one of variables (two in this case) is character mode, cbind coerces Val to character as well. data.frame converts the three character variables to factor (update: as of R 4.0.0 character strings are no longer converted to factors by default). Either way, Val is a discrete (categorical) value rather than numeric, resulting in an error when you use scale_y_continuous.

Change to dfm2 <- data.frame(Group,Statistic,Val) and the error will go away.

You can check the effect of cbind and data.frame on data types as follows:

cbind(Group, Statistic, Val)

     Group Statistic Val   
[1,] "a"   "Mean"    "0.75"
[2,] "b"   "Mean"    "0.83"
...
[7,] "c"   "d"       "0.1" 
[8,] "d"   "d"       "0.3" 

dfm2<-data.frame(cbind(Group,Statistic,Val))
str(dfm2)

'data.frame':   8 obs. of  3 variables:
$ Group    : Factor w/ 4 levels "a","b","c","d": 1 2 3 4 1 2 3 4
$ Statistic: Factor w/ 2 levels "d","Mean": 2 2 2 2 1 1 1 1
$ Val      : Factor w/ 8 levels "0.02","0.1","0.3",..: 6 8 7 5 4 1 2 3

dfm2 <- data.frame(Group,Statistic,Val)
str(dfm2)

'data.frame':   8 obs. of  3 variables:
$ Group    : Factor w/ 4 levels "a","b","c","d": 1 2 3 4 1 2 3 4
$ Statistic: Factor w/ 2 levels "d","Mean": 2 2 2 2 1 1 1 1
$ Val      : num  0.75 0.83 0.79 0.69 0.5 0.02 0.1 0.3

If you don't want data.frame to convert strings to factors, add the argument stringsAsFactors=FALSE.

Hessney

Ah! Thanks--works great now. Even when I was using cbind() incorrectly, I called dfm2$Val<-as.numeric(dfm2$Val) after creating my data frame. str(dfm2) showed $Val as numeric then; but I still got the error. So that is why I was confused. Any ideas on why the as.numeric() adjustment did not work? – Tenet

I don't get an error when I do the as.numeric conversion and then plot it. However, when you convert a factor variable, you need to convert to character first and then to numeric, otherwise you get the factor number codes for the levels, rather than the actual values of the data. Compare dfm2$Val2 = as.numeric(dfm2$Val) and dfm2$Val3 = as.numeric(as.character(dfm2$Val)). – Hessney

Answer 2

Try the following.

ggplot(dfm2,aes(x = Group,y = as.numeric(as.character(Val)))) +    
  geom_bar(aes(fill = Statistic),position = 'dodge',stat='identity')+
  scale_y_continuous(limits=c(0, 1))

Wigan