logreg.tree.control {LogicReg}

Control for logreg

Description

Control of various secondary parameters of tree shape needed in logreg.

Usage

logreg.tree.control(treesize=8, opers=1, minmass=0, n1)

Arguments

treesize

specify the maximum number of allowed leaves per logic tree. Allowing one leave means that the tree is (at most) a simple predictor, two leaves allows for trees such as (X1 or X2) or (not X3 and X4). Four, eight and sixteen leaves allow for two, three or four levels of operators. To be able to interpret the results, do not choose too many leaves. Since the model selection techniques usually trim down the trees, it is recommend to allow at least four or eight leaves per tree.

opers

The default is to allow both "and" and "or" operators in the logic trees. If the interest is in logic statements in disjunctive normal form, use only one of the two operator types. Choose 1 for both operators, 2 for only "and" and 3 for only "or".

minmass

specify the minimum number of cases for which any tree needs to be 1 and for which any tree needs to be 0 to be considered as a logic tree in the model. This is to prevent that logreg, will select trees with, for example, 999 1s and one 0 out of 1000 cases. The default is to take 5% of the cases or 15, whatever is less.

n1

if you specify the sample size n1, it is checked that minmass is smaller than n1/4. This option is used by logreg, but is likely not useful for direct use.

Details

Missing arguments take defaults. If the argument treesize is a list with arguments treesize, opers, and minmass, those values take precedent of directly specified values.

Value

A list with components treesize, opers, and minmass, that can be used as the value of the argument tree.control of logreg.

Author(s)

Ingo Ruczinski [e-mail] and Charles Kooperberg [e-mail].

References

Ruczinski I, Kooperberg C, LeBlanc ML (2003). Logic Regression, Journal of Computational and Graphical Statistics, 12, 475-511.

Ruczinski I, Kooperberg C, LeBlanc ML (2002). Logic Regression - methods and software. Proceedings of the MSRI workshop on Nonlinear Estimation and Classification (Eds: D. Denison, M. Hansen, C. Holmes, B. Mallick, B. Yu), Springer: New York, 333-344.

Selected chapters from the dissertation of Ingo Ruczinski, available from https://research.fredhutch.org/content/dam/stripe/kooperberg/ingophd-logic.pdf

See Also

logreg, logreg.anneal.control, logreg.mc.control

Examples

mytreecontrol <- logreg.tree.control(treesize = 16, minmass = 10)