mscale {pense} | R Documentation |
Compute the M-estimate of scale with MAD as initial estimate.
mscale(x, delta = 0.5, rho = c("bisquare", "huber"), cc, eps = 1e-08, maxit = 200)
x |
numeric vector of observations. |
delta |
target value of the M-estimation equation. |
rho |
rho function to use in the M-estimation equation. |
cc |
non-negative constant for the chosen rho function. If missing,
it will be chosen such that the expected value of the rho function
under the normal model is equal to |
eps |
threshold for convergence. |
maxit |
maximum number of iterations. |
This solves the M-estimation equation given by
∑_{i=1}^n ρ( x_i / s_n; cc ) = n δ
the M-scale as a numeric vector of length one.
NA
values in x
are removed before calculating the
M-scale.
## Estimate the M-scale of a vector of values set.seed(1234) x <- rnorm(100) mscale(x) mscale(x, delta = 0.25) # For a breakdown point of 25% mscale(x, rho = "huber") # Using Huber's rho function