forcing a segmented regression through the origin

seglm_through_origin.R

# forcing a segmented regression through the origin.


# load packages
library(segmented)

# make data
d <- data.frame(x = c(3, 13, 18, 19, 19, 26, 26, 33, 40, 49, 51, 53, 67, 70, 88
),
y = c(3.56211608128595, 10.5214485148819, 3.66063708049802, 6.11000808621074, 
      5.51520423804034, 7.73043895812661, 7.90691392857039, 6.59626527933846, 
      10.4413913666936, 8.71673928545967, 9.93374157928462, 1.214860139929, 
      3.32428882257746, 2.65223361387063, 3.25440939462105))

# fit normal linear regression and segmented regression
lm1 <- lm(y ~ x, d)
lm2 <- lm(y ~ 0 + x, d)
seg_lm <- segmented(lm1, ~ x)
seg_lm2 <- segmented(lm2, ~ x)


slope(seg_lm)

# make predictions
preds1 <- data.frame(x = data.frame(x=0:100), preds = predict(seg_lm, newdata = data.frame(x=0:100)))
preds2 <- data.frame(x = data.frame(x=0:100), preds = predict(seg_lm2, newdata = data.frame(x=0:100)))

# plot segmented fit
plot(seg_lm, res = TRUE, xlim = c(0,95), ylim = c(0,11))

# plot predictions
lines(preds1$preds ~ preds1$x, col = 'blue')
lines(preds2$preds ~ preds2$x, col = 'red')