segreg: Segmented Regression
In JVAdams/jvamisc: A Collection of Utility Functions

Description Usage Arguments Details Value Examples

Fit a simple linear segmented regression (also known as changepoint, hockey stick, or broken line regression).

1	segreg(x, y, k = NULL)

`x`	Numeric vector, the independent variable which will be "broken".
`y`	Numeric vector, the dependent variable which will be used to define the break point.
`k`	Numeric scalar, the location of the break point, if known. If NULL, the default, the optimal break point will be chosen from among all unique values of x except for the minimum and the maximum.

The break point cuts the x vector into two groups, x <= k and x > k.

List with three elements, fit: the resulting lm object, k, and pred: data frame with the unique values of x (and k) with predicted values.

indvar <- sample(1:50, 100, replace=TRUE)
depvar <- ifelse(indvar<32, indvar + 3, indvar/4 + 27) + rnorm(100, sd=2)
sr <- segreg(indvar, depvar)
plot(indvar, depvar)
lines(sr$pred)
abline(v=sr$k, lty=2)