RMM.ZOIP: ZOIP regression with mixed efects

In jucdiaz/ZOIP: ZOIP Distribution, ZOIP Regression, ZOIP Mixed Regression

Description

function RMM.ZOIP fits a mixed ZOIP regression model with random intercepts normals in the mean and dispersion parameter, the estimation is done via maximum likelihood and the gauss-hermite adaptive quadrangle with or without pruning. the model may or may not include effects fixed in any of its parameters, just as it can be a bilaterally inflated model, unilaterally or without parameters inflated.

Usage

RMM.ZOIP(formula.mu, formula.sigma = ~1, formula.p0 = ~1, formula.p1 = ~1,
  data, formula.random, link = c("identity", "identity", "identity",
  "identity"), family = "R-S", optimizer = "nlminb", n.points = 11,
  pruning = TRUE)

Arguments

formula.mu	Formula that defines the regression function for mu, p.e and ~ x1 + x2, it is necessary to define the response variable.
formula.sigma	Formula that defines the regression function for the sigma parameter, a possible value is ~ x1, by default ~ 1.
formula.p0	Formula that defines the regression function for p0, a possible value is ~ x1, by default ~ 1.
formula.p1	Formula that defines the regression function for p1, a possible value is ~ x1, by default ~ 1.
data	It is the data set in data.frame format where it must contain the names of the columns as they are in the formulas.
formula.random	Formula that defines the mixed effect within the model, it should be only the random intercept that will be taken into account in the parameter of the mean and the dispersion, the admissible structure is the following formula.random = ~ 1 | G1, where G1 is the variable that indicates the groups or subjects in the model, should always be defined.
link	It is a vector with the appropriate link functions for each parameter to be estimated according to the options chosen in the family and formula parameters. If the regression model does not have covariables, the identity option should be used as a link function, regardless of the value chosen in the family, possible options are logit, log, default link = c ('identity', 'identity', 'identity', 'identity').
family	choice of the parameterization or distribution, family = 'R-S' parameterization beta distribution Rigby and Stasinopoulos, 'F-C' distribution Beta parametrization Ferrari and Cribari-Neto, 'Original' Beta distribution classic parameterization, 'Simplex' simplex distribution.
optimizer	Choice of the optimizer, used to find the convergence of the maximum likelihood. you can choose the value of 'nlminb' or 'optim', by default 'nlminb'.
n.points	Number of points to use in the approximation of the likelihood function by means of quadrature Gauss-Hermite adaptive multidimensional, by default is 11, it is recommended not to give a very large value to this parameter, because it will significantly affect the times of convergence of the model.
pruning	It is a Boolean value that indicates if pruning is used or not, for the quadrature of multidimensional Adaptive Gauss-Hermite, by default it is TRUE.

Examples

library(ZOIP)

N<-2
ni<-10
set.seed(12345)
Ciudad <- rep(1:N, each=ni)
Total_mora<-rexp(N*ni,rate=1)
set.seed(12345)
b0i <- rep(rnorm(n=N,sd=0.5), each=ni)
set.seed(12345)
b1i <- rep(rnorm(n=N,sd=0.4), each=ni)

neta <- (-1.13+b0i)+0.33*Total_mora
neta2<-(0.33+b1i)+0.14*Total_mora

mu <- 1 / (1 + exp(-neta))
sigma <- 1 / (1 + exp(-neta2))

p0 <- 0.05
p1 <- 0.05

mu[mu==1] <- 0.999
mu[mu==0] <- 0.001

sigma[sigma==1] <- 0.999
sigma[sigma==0] <- 0.001
family<-'R-S'
set.seed(12345)
Y <- rZOIP(n=length(mu), mu = mu, sigma = sigma ,p0=p0,p1=p1,family=family)

data_sim<-data.frame(Y,Total_mora,Ciudad)

n.points <- 3
pruning <- TRUE

formula.mu=Y~Total_mora
formula.sigma=~Total_mora
formula.p0=~1
formula.p1=~1
formula.random= ~ 1 | Ciudad
link=c('logit','logit','identity','identity')
optimizer<-'nlminb'

mod<-RMM.ZOIP(formula.mu=formula.mu,formula.sigma=formula.sigma,formula.p0=formula.p0,
               formula.p1=formula.p1,data=data_sim,formula.random=formula.random,link=link,
               family=family,optimizer=optimizer,n.points=n.points,pruning=pruning)
mod

jucdiaz/ZOIP documentation built on March 17, 2021, 2:11 a.m.