I am experimenting with scatter3d from car package. My code is the following:
library(car)
x <- sample(20 : 100, size = 1000, replace = TRUE)
z <- sample (20:100, size = 1000, replace = TRUE)
y <- (2/3)*x^(3/2) - (z^2)/6 + x * z/3 - 1 /(x/1000 * z/10)
q <- quantile(y, probs=c(0.001, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.999 ))
p <- as.numeric((y > q[2] & y < q[3]) | (y > q[6] & y < q[7]))
scatter3d(p ~ x + z, grid = FALSE, groups = as.factor(p), surface = FALSE)
scatter3d(p ~ x + z, grid = FALSE, fit = "smooth")
The output is the following:
My question is how to interpret the Smooth surface --given that the dependent is binary? Should I interpret it as a probability? However its height is not limited to the range [0,1].
Or perhaps, I should not use a smooth surface with a binary variable?
Your advice will be appreciated.
