Passing through data frames into functions and into ggplot by column

Question

I'm trying to do my first function in R. I have a dataframe of inderminate columns, and I want to create a ggplot of each set of columns. For example, columns, 1&2, 1&3, 1&4 etc.

However, when I try the following function I get the object not found error, but only when we get the the ggplot portion.

Thanks,

BrandPlot=function(Brand){
  NoCol=ncol(Brand)
  count=2
  while (count<=NoCol){
   return(ggplot(Brand, aes(x=Brand[,1], y=Brand[,count]))+geom_point())
    count=(count+1)
  }
}

To clarify,

I'm trying to get the effect (also, I plan on adding additional things like geom_smooth() but I want to get it working first

ggplot(Brand, aes(x=Brand[,1], y=Brand[,2]))+geom_point
ggplot(Brand, aes(x=Brand[,1], y=Brand[,3]))+geom_point
ggplot(Brand, aes(x=Brand[,1], y=Brand[,4]))+geom_point
ggplot(Brand, aes(x=Brand[,1], y=Brand[,5]))+geom_point

(also, I plan on adding additional things like geom_smooth() ) but I want to get it working first

Are you trying to do a function here? It sounds like you may want to melt the data, then add a facet_wrap function — maloneypatr
– maloneypatr, Commented Mar 6, 2014 at 17:24
And note if I switch ggplot to plot, the code works just fine. I'm not sure what I'm doing wrong — Krone
– Krone, Commented Mar 6, 2014 at 17:58

maloneypatr · Accepted Answer · 2014-03-06 18:15:15Z

1

Per the note above, something like this may be what you're looking for...

brandplot <- function(x){
    require(reshape2)
    require(ggplot2)

    x_melt <- melt(x, id.vars = names(x)[1])

    ggplot(x_melt, 
           aes_string(x = names(x_melt)[1], 
                      y = 'value', 
                      group = 'variable')) +
      geom_point() +
      facet_wrap( ~ variable)
}

dat <- data.frame(a = sample(1:10, 25, T),
                  b = sample(20:30, 25, T),
                  c = sample(40:50, 25, T))

brandplot(dat)

enter image description here

edited Mar 6, 2014 at 18:15

answered Mar 6, 2014 at 17:34

maloneypatr

3,6204 gold badges26 silver badges34 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

jlhoward · Accepted Answer · 2014-03-06 18:38:54Z

[Note: @maloneypatr's solution is a better way to use ggplot for your application].

To answer your question directly, there are a couple of problems.

Your function returns after the first run through the loop (e.g., count=2), so you will never get more than one plot from this.
ggplot evaluates arguments to aes(...) in the context of the data frame defined in data=..., so it is looking for something like Brand$Brand (e.g., a column named Brand in the dataframe Brand). Since there is no such column, you get the Object not found error.

The following code will generate a series of n-1 plots where n = ncol(Brand).

BrandPlot=function(Brand){
  for (count in 2:ncol(Brand)){
    ggp <- ggplot(Brand, aes_string(x=names(Brand)[1], y=names(Brand)[count]))
    ggp <- ggp + geom_point()
    ggp <- ggp + ggtitle(paste(names(Brand)[count], " vs. ", names(Brand)[1]))
    plot(ggp)
  }
}

So I need to use the "names" function for it to know. Got it. Thanks so much. I'm curious though why I can call it that way outside of a function. And it works with regular plot.
I looked into this some more, and it appears that ggplot does not recognize the variable count defined in the function. However, if you define count in the calling environment, ggplot finds it. So it looks like ggplot first looks for a column count, as in Brand$count, and not finding that seeks a count variable in the environment from which the function was called. Very strange. Bottom line: if at all possible, only use column names in aes(...). The plot(...) function in base R works completely differently. There is no equivalent to aes(...).

Collectives™ on Stack Overflow

Passing through data frames into functions and into ggplot by column

2 Answers 2

Comments

2 Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Related