Question

I have been unable to find a simple analog for plotting a line graph from a table object in ggplot2. Given the elegance and utility of the package, I feel I must be missing something quite obvious. As an illustration consider a data frame with yearly observations:

dat<-data.frame(year=sample(c("2001":"2010"),1000, replace=T))

And a quick time series plot in base R:

plot(table(dat$year), type="l")

Switching to qplot, returns the error “attempt to apply a non-function”:

qplot(table(dat$year), geom="line")

ggplot2 requires a data frame. Fair enough. But this returns the same error.

qplot(year, data=dat, geom="line")

After some searching and fiddling, I abandoned qplot, and came up with the following approach which involves specifying a line geometry, binning the counts, and dropping final values to avoid plotting zeros.

ggplot(dat, aes(year) ) + geom_line(stat = "bin", binwidth=1, drop=TRUE)

It seems like rather a long walk around the block. And it is still not entirely satisfactory, since the bins don’t align precisely with the mid-year values on the x-axis. Where have I gone wrong?

Answer 1

Maybe still more complicated than you want, but:

qplot(Var1,Freq,data=as.data.frame(table(dat$year)),geom="line",group=1)

(the group=1 is necessary because the Year variable (Var1) is returned as a factor …)

If you didn’t need it as a one-liner you could use ytab <- as.data.frame(table(dat$year)) first to extract the table and convert it to a data frame …

Following Brian Diggs’s answer, if you’re willing to construct a bit more fortify machinery you can condense this a bit more:

A utility function that converts a factor to numeric if possible:

conv2num <- function(x) {
    xn <- suppressWarnings(as.numeric(as.character(x)))
    if (!all(is.na(xn))) xn else x
}

And a fortify method that turns the table into a data frame and then tries to make the columns numeric:

fortify.table <- function(x,...) {
    z <- as.data.frame(x)
    facs <- sapply(z,is.factor)
    z[facs] <- lapply(z[facs],conv2num)
    z
}

Now this works almost as you would like it to:

qplot(Var1,Freq,data=table(dat$year),geom="line")

(It would be nice/easier if there were a table option to preserve the numeric nature of cross-classifying factors …)