How to mark slope changes in LOESS curve using ggplot2?

Question

I have some time-series data that I\'m fitting a loess curve in ggplot2, as seen attached. The data takes the shape of an \"S\" curve. What I really need to find out is the

Answer 1

If you are asking for a way of determining the point where the curve is a maximum (i.e. flat), this is the same as finding the point where the slope of the line is at its maximum (from basic calculus).

First, read your data:

christi <- read.table("http://dl.dropbox.com/u/75403/stover_data.txt", sep="\t", header=TRUE)

Next, use loess to fit a smoothed model:

fit <- loess(org_count~date, data=christi)

Then, predict the values in your range of x-values (with predict.loess), determine the slope (diff is close enough`), and find the

x <- 200:800
px <- predict(fit, newdata=x)
px1 <- diff(px)

which.max(px1)
[1] 367

Since the start value of x is 200, this means the curve is flat at position 200+367=567.

---

If you wanted to plot this:

par(mfrow=c(1, 2))
plot(x, px, main="loess model")

plot(x[-1], px1, main="diff(loess model)")
abline(v=567, col="red")