COUNT (version 1.3.4)

nb2_syn: Negative binomial (NB2): generic synthetic negative binomial data and model

Description

nb2_syn is a generic function for developing synthetic NB2 data and a model given user defined specifications.

Usage

nb2_syn(nobs = 50000, off = 0, alpha = 1, xv = c(1, 0.75, -1.5))

Value

nby

Negative binomial response; number of counts

sim.data

synthetic data set

Arguments

nobs

number of observations in model, Default is 50000

alpha

NB2 heterogeneity or ancillary parameter

off

optional: log of offset variable

xv

predictor coefficient values. First argument is intercept. Use as xv = c(intercept , x1_coef, x2_coef, ...)

Author

Andrew Robinson, Universty of Melbourne, Australia, and Joseph M. Hilbe, Arizona State University, Jet Propulsion Laboratory, California Institute of Technology

Details

Create a synthetic negative binomial (NB2) regression model using the appropriate arguments. Model data with predictors indicated as a group with a period (.). Offset optional. If no offset is desired, drop "off= loff" from nb2_syn function call and "+ loff" from glm.nb function call. See examples.

Data can be estimated using the glm.nb() function, or the ml.nb2() function in the COUNT package, or by using the gamlss function in the gamlss package, with "family=NBI" option.

References

Hilbe, J.M. (2011), Negative Binomial Regression, second edition, Cambridge University Press.

See Also

poisson_syn, nb1_syn, nbc_syn

Examples

library(MASS)           

sim.data <- nb2_syn(nobs = 500, alpha = .5, xv = c(2, .75, -1.25))
mynb2 <- glm.nb(nby ~ . , data = sim.data)
summary(mynb2)
confint(mynb2)

# with offset
oset <- rep(1:5, each=100, times=1)*100 
loff <- log(oset)   
sim.data <- nb2_syn(nobs = 500, off = loff, alpha = .5, xv = c(1.2, -.75, .25, -1.3))
mypof <- glm.nb(nby ~ . + loff, data = sim.data)
summary(mypof)
confint(mypof)

# without offset, exponentiated coefficients, CI's
sim.data <- nb2_syn(nobs = 500, alpha = .75, xv = c(1, .5, -1.4))
mynbf <- glm.nb(nby ~ . , data = sim.data)
exp(coef(mynbf))
exp(confint(mynbf))

if (FALSE) {
# default, without offset
sim.data <- nb2_syn()
dnb2 <- glm.nb(nby ~ . , data = sim.data)
summary(dnb2)
}

# use ml.nb2.r function
sim.data <- nb2_syn(nobs = 500, alpha = .5, xv = c(2, .75, -1.25))
mynb2x <- ml.nb2(nby ~ . , data = sim.data)
mynb2x

if (FALSE) {
# use gamlss function for modeling data after sim.data created
library(gamlss)
sim.data <- nb2_syn(nobs = 500, alpha = .5, xv = c(2, .75, -1.25))
gamnb <- gamlss(nby ~ ., family=NBI, data = sim.data)
gamnb
}