meteorits (version 0.1.1)

emSNMoE: emSNMoE implements the ECM algorithm to fit a Skew-Normal Mixture of Experts (SNMoE).

Description

emSNMoE implements the maximum-likelihood parameter estimation of a Skew-Normal Mixture of Experts (SNMoE) model by the Expectation Conditional Maximization (ECM) algorithm.

Usage

emSNMoE(X, Y, K, p = 3, q = 1, n_tries = 1, max_iter = 1500,
  threshold = 1e-06, verbose = FALSE, verbose_IRLS = FALSE)

Arguments

X

Numeric vector of length n representing the covariates/inputs \(x_{1},\dots,x_{n}\).

Y

Numeric vector of length n representing the observed response/output \(y_{1},\dots,y_{n}\).

K

The number of experts.

p

Optional. The order of the polynomial regression for the experts.

q

Optional. The order of the logistic regression for the gating network.

n_tries

Optional. Number of runs of the ECM algorithm. The solution providing the highest log-likelihood will be returned.

max_iter

Optional. The maximum number of iterations for the ECM algorithm.

threshold

Optional. A numeric value specifying the threshold for the relative difference of log-likelihood between two steps of the ECM as stopping criteria.

verbose

Optional. A logical value indicating whether or not values of the log-likelihood should be printed during ECM iterations.

verbose_IRLS

Optional. A logical value indicating whether or not values of the criterion optimized by IRLS should be printed at each step of the ECM algorithm.

Value

ECM returns an object of class ModelSNMoE.

Details

emSNMoE function implements the ECM algorithm for the SNMoE model. This function starts with an initialization of the parameters done by the method initParam of the class ParamSNMoE, then it alternates between the E-Step (method of the class StatSNMoE) and the M-Step (method of the class ParamSNMoE) until convergence (until the relative variation of log-likelihood between two steps of the ECM algorithm is less than the threshold parameter).

See Also

ModelSNMoE, ParamSNMoE, StatSNMoE

Examples

# NOT RUN {
data(tempanomalies)
x <- tempanomalies$Year
y <- tempanomalies$AnnualAnomaly

snmoe <- emSNMoE(X = x, Y = y, K = 2, p = 1, verbose = TRUE)

snmoe$summary()

snmoe$plot()
# }