twinstim_step: Stepwise Model Selection by AIC

In jimhester/surveillance: Temporal and Spatio-Temporal Modeling and Monitoring of Epidemic Phenomena

Description

stepComponent is a wrapper around step to select a "twinstim" component's model based on an information criterion in a stepwise algorithm.

There are also stand-alone single-step methods of add1 and drop1.

Usage

stepComponent(object, component = c("endemic", "epidemic"),
              scope = list(upper = object$formula[[component]]),
              direction = "both", trace = 2, verbose = FALSE, ...)

## S3 method for class 'twinstim'
add1(object, scope, component = c("endemic", "epidemic"), 
    trace = 2, ...)
## S3 method for class 'twinstim'
drop1(object, scope, component = c("endemic", "epidemic"), 
     trace = 2, ...)

Arguments

object	an object of class "twinstim".
component	one of "endemic" or "epidemic" (partially matched), determining the model component where the algorithm should proceed.
scope,direction,trace	see step and add1, respectively.
verbose	see twinstim.
...	further arguments passed to step, add1.default, or drop1.default, respectively.

Value

See step and add1, respectively.

Author(s)

(of this wrapper around step) Sebastian Meyer

See Also

step, add1, drop1

Examples

data("imdepi")
data("imdepifit")

## simple baseline model
m0 <- update(imdepifit, epidemic=~1, siaf=NULL, start=c("e.(Intercept)"=-17))

## AIC-based step-wise backward selection of the endemic component
m0_step <- stepComponent(m0, "endemic", scope=list(lower=~I(start/365-3.5)))
## nothing is dropped from the model

jimhester/surveillance documentation built on May 19, 2019, 10:33 a.m.