Summarize GLM Model Selection via Branch and Bound

Share:

Description

These functions are all methods for class glmbb or summary.glmbb objects.

Usage

1
2
3
4
5
6
## S3 method for class 'glmbb'
summary(object, cutoff, ...)

## S3 method for class 'summary.glmbb'
print(x, digits = max(3, getOption("digits") - 3),
    ...)

Arguments

object

an object of class "glmbb", usually, a result of a call to glmbb.

cutoff

a nonnegative real number. Only report on models having criterion value no larger than the minimum value plus cutoff. This argument may be omitted, in which case object$cutoff is used.

x

an object of class "summary.glmbb", usually, a result of a call to summary.glmbb.

digits

the number of significant digits to use when printing.

...

not used. Required by their generics.

Details

Let criterion denote the vector of criterion (AIC, BIC, or AICc) values for all of the models evaluated in the search. Those with criterion value greater than min(criterion) + cutoff are tossed.

We also define a vector weight by

1
2
w <- exp(- criterion / 2)
weight <- w / sum(w)

except that it is calculated differently to avoid overflow. These are so-called Akaike weights. They may or may not provide some guide as to how to deal with these models. For more see Burnham and Anderson (2002).

Value

summary.glmbb returns an object of class "summary.glmbb", a list with components

results

a data frame having variables

criterion

the vector criterion described in the Details section, in sorted order.

weight

the corresponding Akaike weights.

formula

the corresponding formulas describing the corresponding models.

cutoff.search

the cutoff argument to the call to glmbb that produced object.

cutoff.summary

the cutoff argument to the call to summary.glmbb.

criterion

a character variable giving the name of the criterion (AIC, BIC, or AICc). Not to be confused with results$criterion.

References

Burnham, K. P. and Anderson, D. R. (2002). Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd ed. Springer-Verlag, New York.

Examples

1
## For examples see those in help(glmbb)