Small Dataset Example of He
The exHe data frame has 10 rows and 2 columns. It is an
example for which smooth.spline cannot be used.
data(exHe)
This data frame contains the following columns:
only values 0, 1, and 2.
10 randomly generated values
Xuming He wrote about this
JUST FOR FUN:
I was testing COBS using the following "data". For comparison, I tried
smooth.spline in S+. I never got an answer back! No warning messages either.
The point is that even the well-tested algorithm like
smooth.spline could leave you puzzled.
To tell you the truth, the response values here were generated by white noise. An ideal fitted curve would be a flat line. See for yourself what COBS would do in this case.
Originally found at the bottom of
http://ux6.cso.uiuc.edu/~x-he/ftp.html, the web resource
directory of Xuming He at the time, say 2006.
data(exHe)
plot(exHe, main = "He's 10 point example and cobs() fits")
tm <- tapply(exHe$y, exHe$x, mean)
lines(unique(exHe$x), tm, lty = 2)
cH. <- with(exHe,
cobs(x, y, degree=1, constraint = "increase"))
cH <- with(exHe,
cobs(x, y, lambda=0.2, degree=1, constraint = "increase"))
plot(exHe)
lines(predict(cH.), type = "o", col="tomato3", pch = "i")# constant
lines(predict(cH), type = "o", col=2, pch = "i")
cHn <- cobs(exHe$x, exHe$y, degree=1, constraint = "none")
lines(predict(cHn), col= 3, type = "o", pch = "n")
cHd <- cobs(exHe$x, exHe$y, degree=1, constraint = "decrease")
lines(predict(cHd), col= 4, type = "o", pch = "d")Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.