R ggplot2 introduce slight smoothing to a line graph with only a few datapoints

Question

Not sure if this is a programming question or not...

If I have the data below, which produces a \'spiky\' chart, and I\'d like to produce a slightly smoothed one using g

Answer 1

If you want to avoid losing too much information from the data, below could be a better approach, that works good for large datasets:

library(zoo)
library(reshape)
a$smooth<-rollmean(a$values,3,fill="extend") # 2nd parameter defines smoothness 
ggplot(melt(a),aes(x=year,y=value,color=variable,group=variable))+geom_line()

Here is a better example:

a <- data.frame(year=1:10,values=sin(1:10)+runif(10))
a$smooth<-rollmean(a$values,3,fill="extend")
ggplot(melt(a,id.vars="year"),aes(x=year,y=value,color=variable,
      group=variable))+geom_line(size=2)

Answer 2

You can try a polynomial. Since the x-axis variable has 12 unique values, you can use polynomials up to the 11th degree. Furthermore, you should use a continuos scale for the x-axis to achieve a smooth curve.

Here's an example of an 8th-order polynomial:

ggplot(a, aes(x = year, y = values, group = 1))+
  stat_smooth(aes(x = seq(length(unique(year)))), # continuous x-axis
              se = F, method = "lm", formula = y ~ poly(x, 8)) +
  scale_x_continuous(breaks = seq(length(unique(a$year))), 
                     labels = levels(a$year)) # original labels

Here, method = "lm" means, that a linear model is used. The second argument of the poly function specifies the degree.