update_eta {EMSNM} | R Documentation |
Updata eta in step t+1 with given data and coeffients estimated in step t.
update_eta(fun, alphat, sigmat, etat, X, Y, Z, learning_rate_eta = 0.001, regular_parameter_eta = 0.001, max_iteration_eta = 10000)
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. |
alpha |
alpha estimated in step t. |
eta |
eta estimated in step t+1. |
sigma |
sigma estimated in step t. |
Linsui Deng
#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)