stepAIC {MASS} | R Documentation |
Performs stepwise model selection by exact AIC.
stepAIC(object, scope, scale, direction=c("both", "backward", "forward"), trace=1, keep=NULL, steps=1000, use.start=FALSE, k=2, ...) extractAIC(fit, scale, k=2, ...)
object fit |
an object representing a model of an appropriate class. This is used as the initial model in the stepwise search. |
scope |
defines the range of models examined in the stepwise search. |
scale |
used in the definition of the AIC statistic for selecting the models,
currently only for lm , aov and glm models.
|
direction |
the mode of stepwise search, can be one of "both" , "backward" , or
"forward" , with a default of "both" . If the scope argument is
missing, the default for direction is "backward" .
|
trace |
if positive, information is printed during the running of stepAIC() .
Larger values may give more information on the fitting process.
|
keep |
a filter function whose input is a fitted model object and the
associated AIC statistic, and whose output is arbitrary.
Typically keep will select a subset of the components of
the object and return them. The default is not to keep anything.
|
steps |
the maximum number of steps to be considered. The default is 1000 (essentially as many as required). It is typically used to stop the process early. |
use.start |
if true the updated fits are done starting at the linear predictor for
the currently selected model. This may speed up the iterative
calculations for glm (and other fits), but it can also slow them
down.
|
k |
the multiple of the number of degrees of freedom used for the penalty.
Only k=2 gives the genuine AIC: k = log(n) is sometimes referred
to as BIC or SBC.
|
... |
any additional arguments to extractAIC . (None are currently used.)
|
stepAIC
differs from step
and especially step.glm
in using the
exact AIC rather than potentially misleading one-step approximations.
It is also much more widely applicable: all that is required is a
method for extractAIC
, which should return a vector
c(modeldf, AIC)
.
The default method handles linear models (lm
, aov
and
glm
of family "Gaussian"
with identity link) using addterm.lm
and
dropterm.lm
: for these the results are similar to step.glm
except
that the AIC quoted is Akaike's not Hastie's. (The additive constant
is chosen so that in that case AIC is identical to Mallows' Cp if the
scale is known.)
There is a potential problem in using glm
fits with a variable
scale
, as in that case the deviance is not simply related to the
maximized log-likelihood. The function extractAIC.glm
makes the
appropriate adjustment for a gaussian
family, but may need to be
amended for other cases. (The binomial
and poisson
families have
fixed scale
by default and do not correspond to a particular
maximum-likelihood problem for variable scale
.)
Where a conventional deviance exists (e.g. for lm
, aov
and glm
fits) this is quoted in the analysis of variance table: it is the
unscaled deviance.
the stepwise-selected model is returned, with up to two additional
components. There is an "anova"
component corresponding to the
steps taken in the search, as well as a "keep"
component if the
keep=
argument was supplied in the call. The "Resid. Dev"
column
of the analysis of deviance table refers to a constant minus twice the
maximized log likelihood: it will be a deviance only in cases where a
saturated model is well-defined (thus excluding lm
, aov
and
survreg
fits, for example).
data(quine) quine.hi <- aov(log(Days + 2.5) ~ .^4, quine) quine.nxt <- update(quine.hi, . ~ . - Eth:Sex:Age:Lrn) quine.stp <- stepAIC(quine.nxt, scope = list(upper = ~Eth*Sex*Age*Lrn, lower = ~1), trace = FALSE) quine.stp$anova data(cpus) cpus1 <- cpus attach(cpus) for(v in names(cpus)[2:6]) cpus1[[v]] <- cut(cpus[[v]], unique(quantile(cpus[[v]])), include.lowest = TRUE) detach() cpus0 <- cpus1[, 2:8] # excludes names, authors' predictions cpus.samp <- sample(1:209, 100) cpus.lm <- lm(log10(perf) ~ ., data=cpus1[cpus.samp,2:8]) cpus.lm2 <- stepAIC(cpus.lm, trace=FALSE) cpus.lm2$anova example(birthwt) birthwt.glm <- glm(low ~ ., family=binomial, data=bwt) birthwt.step <- stepAIC(birthwt.glm, trace=FALSE) birthwt.step$anova birthwt.step2 <- stepAIC(birthwt.glm, ~ .^2 + I(scale(age)^2) + I(scale(lwt)^2), trace=FALSE) birthwt.step2$anova quine.nb <- glm.nb(Days ~ .^4, data=quine) quine.nb2 <- stepAIC(quine.nb) quine.nb2$anova