logit P(Y <= j) = beta_j0 + x'beta fitted by Newton-Raphson with the analytic
Hessian and step halving. MASS::polr reports zeta = beta_j0 and -beta.
Usage
PolrFit(y, X, max_iter = 200, tol = 1e-12)
Arguments
- y
Ordered categories coded 0, ..., J - 1 (all present).
- X
Matrix of explanatory variables without intercept.
- max_iter, tol
Newton controls.
Value
list(intercepts, beta, se_intercepts, se_beta, loglik, iterations).
References
Bilder, C. R. & Loughin, T. M. (2025). Analysis of Categorical
Data with R, 2nd ed. Eqs (3.11), (3.15).
Examples
PolrFit(c(0, 0, 1, 0, 1, 2, 1, 2, 2, 1), matrix(0:9))$beta
#> [1] -0.6724748
