update_eta: Updata Eta

In EMSNM: EM Algorithm for Sigmoid Normal Model

Description

Updata eta in step t+1 with given data and coeffients estimated in step t.

Usage

update_eta(fun, alphat, sigmat, etat, X, Y, Z, learning_rate_eta = 0.001, 
           regular_parameter_eta = 0.001, max_iteration_eta = 10000)

Arguments

fun	the function updata eta
alphat	the estimated coeffients of the mean of each subgroup in step t
sigmat	the estimated standard error of Y in step t
etat	the estimated coeffients determining subgroup in step t
X	the covariables of the mean of each subgroup
Z	the covaraibles determining subgroup
Y	the respond variable
learning_rate_eta	learning rate of updating eta
regular_parameter_eta	regular value of updating eta by gradiant descending methond.
max_iteration_eta	maximal steps of eta interation to avoid unlimited looping.

Value

alpha	alpha estimated in step t.
eta	eta estimated in step t+1.
sigma	sigma estimated in step t.

Author(s)

Linsui Deng

Examples

#some variables
samplesize <- 1000
classsize <- 6
etasize <- 3
alphasize <- 2

Xtest <- data.frame(matrix(rnorm(samplesize*etasize),samplesize,etasize))
Ztest <- matrix(rnorm(samplesize*alphasize),samplesize,alphasize)

etatest <- matrix(seq(1.15,1,length=etasize*classsize),etasize,classsize)
alphatest <- matrix(seq(1.15,1,length=alphasize*classsize),alphasize,classsize)
sigmatest <- 0.1

Wtest <- Wgenerate(alpha=alphatest,eta=etatest,X=Xtest,Z=Ztest)

#test of update_eta
thetaupdate_eta <- update_eta(fun=eta_gradient_fun,alphat=alphatest,sigmat=sigmatest,
           etat=etatest,X=Wtest$X,Z=Wtest$Z,Y=Wtest$Y,
           learning_rate=0.1,regular_parameter=0.001,max_iteration=10000)

EMSNM documentation built on May 2, 2019, 1:41 p.m.