initial {glarma}

Initial Parameter Generator for GLARMA from GLM

Description

Function used to generate initial values of parameters for the GLARMA model from glm or glm.nb.

Usage

initial(y, X, offset = NULL, type = "Poi", alpha = 1)

Arguments

y

Numeric vector; response variable.

X

Matrix; explanatory variables. A vector of ones should be added to the data matrix as the first column for the \beta of the intercept.

offset

Either NULL or a numeric vector of length equal to the number of cases. Used to specify an a priori known component to be included in the linear predictor during fitting.

type

Character; the distribution of the counts.

alpha

Numeric; an optional initial value for the theta parameter in the negative binomial distribution; the default value is 1.

Details

Generates and returns the initial parameters for the GLARMA model under the specified distribution by fitting a generalized linear model.

Value

beta

A named numeric vector of initial coefficients.

y

If requested, the y vector used.

X

If requested, the model matrix.

alpha

The theta parameter in the negative binomial distribution returned by glm.nb. NULL if any other distribution is used.

type

The distribution of the counts in the GLARMA model.

null.deviance

Null deviance of the GLM with the same regression structure as the GLARMA model.

df.null

Null degrees of freedom of the GLM with the same regression structure as the GLARMA model.

Author(s)

"William T.M. Dunsmuir" <[e-mail]>

Examples

### Using the polio data
data(Polio)
y <- Polio[, 2]
X <- as.matrix(Polio[, 3:8])

glmMod <- initial(y, X, type = "Poi", alpha=1)
str(glmMod)
head(glmMod)