Incorrect Legend Ordering with scale_color_grey()

Question

I am doing a Monte Carlo simulation, in which I have to display the density of coefficient estimates for simulations with different sample sizes on the same plot. When using scale_color_grey . I have put my coefficient estimates in the same dataframe, with the sample size as a factor. If I query the factor with levels() , it is in the correct order (from smallest to highest sample size). However, the following code gives a scale in which the order is correct in the legend, but the color moves from light grey to darker grey in a seemingly random order

montecarlo <- function(N, nsims, nsamp){
  set.seed(8675309)
  coef.mc <- vector()
    for(i in 1:nsims){
      access <- rnorm(N, 0, 1) 
      health <- rnorm(N, 0, 1)
      doctorpop <- (access*1) + rnorm(N, 0, 1)
      sick <- (health*-0.4) + rnorm(N, 0, 1)
      insurance <- (access*1) + (health*1) + rnorm(N, 0, 1)
      healthcare <- (insurance*1) + (doctorpop*1) + (sick*1) + rnorm(N, 0, 1)
      data <- as.data.frame(cbind(healthcare, insurance, sick, doctorpop))
      sample.data <- data[sample(nrow(data), nsamp), ]
      model <- lm(data=sample.data, healthcare ~ insurance + sick + doctorpop)
      coef.mc[i] <- coef(model)["insurance"]
    }
  return(as.data.frame(cbind(coef.mc, nsamp)))
}

sample30.df <- montecarlo(N=1000, nsims=1000, nsamp=30)
sample100.df <- montecarlo(1000,1000,100)
sample200.df <- montecarlo(1000, 1000, 200)
sample500.df <- montecarlo(1000, 1000, 500)
sample1000.df <- montecarlo(1000, 1000, 1000)
montecarlo.df <- rbind(sample30.df, sample100.df, sample200.df, sample500.df, sample1000.df)
montecarlo.df$nsamp <- as.factor(montecarlo.df$nsamp)
levels(montecarlo.df$nsamp) <- c("30", "100", "200", "500", "1000")

##creating the plot
montecarlo.plot <- ggplot(data=montecarlo.df, aes(x=coef.mc, color=nsamp))+
  geom_line(data = subset(montecarlo.df, nsamp==30), stat="density")+
  geom_line(data = subset(montecarlo.df, nsamp==100), stat="density")+
  geom_line(data = subset(montecarlo.df, nsamp==200), stat="density")+
  geom_line(data = subset(montecarlo.df, nsamp==500), stat="density")+
  geom_line(data = subset(montecarlo.df, nsamp==1000), stat="density")+
  scale_color_grey(breaks=c("30", "100","200", "500", "1000"))+
  labs(x=NULL, y="Density of Coefficient Estimate: Insurance", color="Sample Size")+
  theme_bw()
montecarlo.plot 

Not using the breaks argument to scale_color_grey returns a legend in which the shades are in the right order, but which does not increase from smallest to highest sample size.

What is going on here? As far as I understand it, ggplot2 should follow the factor's order (which is correct) in assigning colors and creating the legend. How can I make both the legend and the shades of grey increase from smallest to lowest sample size?

Answer 1

You should let ggplot handle drawing the separate lines for each level of nsamp : because you have mapped nsamp to the colour aesthetic, ggplot will automatically draw a different line for each level, so you can do:

montecarlo.plot <- ggplot(data=montecarlo.df, aes(x=coef.mc, color=nsamp))+
    geom_line(stat = "density", size = 1.2) +
    scale_color_grey() +
    labs(x=NULL, y="Density of Coefficient Estimate: Insurance", color="Sample Size")+
    theme_bw()
montecarlo.plot

No need to manually subset the data.