newton_raphson_mstep {multinomialLogitMix}

M-step of the EM algorithm

Description

Implements the maximization step of the EM algorithm based on a ridge-stabilized version of the Newton-Raphson algorithm, see Goldfeld et al. (1966).

Usage

newton_raphson_mstep(y, X, b, w, maxNR = 5, R0 = 0.1, method = 5, verbose = FALSE)

Arguments

y

count data matrix

X

design matrix (including const).

b

coefficients of the multinomial logit mixture

w

mixing proportions

maxNR

threshold

R0

inital value for the parameter that controls the step-size of the update.

method

set to 5. Always.

verbose

Boolean.

Value

b

coefficients

theta

theta values

ll

log-likelihood.

Author(s)

Panagiotis Papastamoulis

References

Goldfeld, S. M., Quandt, R. E., and Trotter, H. F. (1966). Maximization by quadratic hill-climbing. Econometrica: Journal of the Econometric Society, 541-551.