ctqr.control: Auxiliary Function for Root Search
In ctqr: Censored and Truncated Quantile Regression

Description Usage Arguments Details Value See Also

View source: R/A2_aux_functions.R

This functions can be used within a call to ctqr, to control the operational parameters of the root search algorithm.

1	ctqr.control(tol = 1e-06, maxit = 1000, a = 0.5, b = 1.25)

`tol`	positive convergence tolerance: the algorithm stops when the maximum absolute change between two consecutive estimates is smaller than `tol`.
`maxit`	maximum number of iterations.
`a,b`	numeric scalar with `0 < a < 1` and `b > 1`. See ‘Details’.

For a current estimate beta, a new estimate is computed as beta_new = beta + delta*s(beta), where s(beta) is the current value of the estimating equation and delta is a positive multiplier. If sum(s(beta_new)^2) < sum(s(beta)^2), the iteration is accepted and delta is multiplied by b. Otherwise, beta_new is rejected and delta is multiplied by a. By default, a = 0.5 and b = 1.25. Choosing a,b closer to 1 may result in a more accurate estimate, but will require a larger number of iterations.