How to plot weighted loess smoothing in ggplot2?

Question

How do I add a loess-smoothing which respects another column as weights?

Let's say I have the following data.frame:

    library(ggplot2)

    df <- data.frame(x=seq(1:21))
    df$y <- df$x*0.3 + 10
    df$weight <- 10

    df[6,] <- c(6, 0.1, 1)
    df[7,] <- c(7, 0.1, 1)

    df[13,] <- c(13, 0.1, 1)
    df[14,] <- c(14, 0.1, 1)

    df[20,] <- c(20, 0.1, 1)
    df[21,] <- c(21, 0.1, 1)

    ggplot(data=df, aes(x=x, y=y, size=weight)) + geom_point() + geom_smooth(method=loess, legend=FALSE)

Plotting a loess-smoothing brings up the following:

But I want to use the column weight such that loess is the same as if every point with weight 10 is 10-times existent:

    df2 <- subset(df, !(x %in% c(6,7,13,14,20,21)))
    df2$weight <- 1

    df3 <- df2
    for(i in seq(1:9)){
      df3 <- rbind(df3, df2)
    }
    df3 <- rbind(df3, c(6, 0.1, 1))
    df3 <- rbind(df3, c(7, 0.1, 1))

    df3 <- rbind(df3, c(13, 0.1, 1))
    df3 <- rbind(df3, c(14, 0.1, 1))

    df3 <- rbind(df3, c(20, 0.1, 1))
    df3 <- rbind(df3, c(21, 0.1, 1))

    ggplot(data=df3, aes(x=x, y=y)) + geom_point() + geom_smooth(method=loess, legend=FALSE) 
    # or to demonstrate
    ggplot(data=df3, aes(x=x, y=y)) + geom_jitter() + geom_smooth(method=loess, legend=FALSE)

I found the parameter weights of the function loess. But I don't know how to invoke it from geom_smooth().

Answer 1

I guess I have to add only another parameter to aes:

  ggplot(data=df, aes(x=x, y=y, size=weight, weight=weight)) + geom_point() + geom_smooth(method=loess, legend=FALSE)