The arboretum procedure

MAXSTEP=n

Specifies the maximum number of steps for the STEPWISE variable selection method.

Default:

Two times the number of effects specified in the MODEL statement.

NODESIGNPRINT

Suppresses the display of the coding of the CLASS inputs.

ALIAS:

NODP

RULE=MULTIPLE | SINGLE | NONE

Specifies the rule for inclusion of effects for SELECTION=FORWARD, BACKWARD, or

STEPWISE.

MULTIPLE

One or more effects can be considered for entry or removal at the same time provided the

hierarchical rule is observed. For example, if main effects A and B and interactions A*B are

not in the model, effects that can be considered for entry in a single step are A alone, or B

alone, or A, B, and A*B together.

SINGLE

A single effect is considered for entry into the model only if its lower order effects are

already in the model; a single effect is considered for removal from the model only if its

higher order effects are not in the model.

NONE

Effects are included or excluded one at a time without preservation of any hierarchical

order.

Default:

RULE=NONE

Reference:

For more complete definitions of the selection methods, see pages 1397-1398

of the "SAS/STAT User's Guide, Volume 2, GLM-VARCOMP, Version 6

Edition."

SELECTION= FORWARD | BACKWARD | STEPWISE | NONE

Specifies the variable selection methods.

FORWARD

Begins with no inputs in the model and then, systematically, adds inputs that are related to

the target.

BACKWARD

Begins with all inputs in the model and then, systematically, removes inputs that are not

related to the target.

STEPWISE

Systematically adds and deletes inputs from the model. Stepwise selection is similar to

forward selection except that stepwise may remove an input after it has entered the model

---

and replace it with another input.

NONE

All inputs are used to fit the model.

Default:

NONE

SEQUENTIAL

Specifies the addition or deletion of variables in sequential order, as specified in the MODEL

statement.

SLENTRY=n

Specifies the significance level for addition of variables.

Default:

.05

SLSTAY=n

Specifies the significance level for removal of variables.

Default:

.05

START=n

Specifies that the first n effects be included in the starting model.

Default:

0 - for the FORWARD or the STEPWISE method

s (the total number of effects in the MODEL statement)- for the

BACKWARD method

Range:

The value of n ranges from 0 to s, where s is the total number of effects in the

MODEL statement.

STOP=n

Specifies the maximum (FORWARD method) or minimum (BACKWARD method) number of

effects to be included in the final model. The variable selection process is stopped when n effects

are added or deleted. The STOP= option has no effect when SELECTION=NONE or STEPWISE.

Range:

The value of n ranges from 0 to s, where s is the total number of effects in the

MODEL statement.

Default:

s - for the FORWARD method

0 - for the BACKWARD method

MODEL Options - Specification Options

---

CODING= DEVIATION | GLM

Specifies design variable coding for CLASS inputs.

DEVIATION

Deviation from mean coding, which is also known as effect coding.

GLM

Non-full rank GLM coding as usedin the GLM procedure.

Default:

CODING=DEVIATION

LEVEL=INTERVAL | NOMINAL | ORDINAL

Specifies the measurement level of the target variable.

INTERVAL

Interval variable.

NOMINAL

Nominal variable.

ORDINAL

Ordinal variable.

Default:

ORDINAL for a categorical target; INTERVAL for a numerical target.

ERROR=MBERNOULLI | NORMAL

Specifies the error distribution.

MBERNOULLI

Multinomial distribution with on trial. This includes the binomial distribution with on trial.

MBERNOULLI is not available if the target meausurement level is interval.

Alias:

BINOMAIL or MULTINOMIAL

NORMAL

Normal distribution. NORMAL is not allowed e if the target measurement level is nominal.

Default:

ERROR=NORMAL (for LEVEL=INTERVAL), ERROR=MBERNOULLI

(otherwise).

LINK= CLOGLOG | IDENTITY | LOGIT | PROBIT

Specifies the link function that represents the expected values of the target to the linear predictors.

CLOGLOG

Specifies the complementary log-log function, which is the inverse of the extreme value

distribution function. The CLOGLOG function is available is available for ordinal or binary

targets.

IDENTITY

Specifies the identity function. The IDENTITY function can only be used for the linear

---

regression analysis (ERROR=NORMAL).

LOGIT

Specifies the logit function, which is the inverse of the logistic distribution function. The

LOGIT function is available for nominal, ordinal, or binary targets.

PROBIT

Specifies the probit function, which is the inverse of the standard normal distribution

function. The PROBIT function is available is available for ordinal or binary targets.

Default:

LOGIT (for ERROR=MBERNOULLI), IDENTITY (for

ERROR=NORMAL).

IDENTITY (for ERROR=NORMAL)

Tip:

The CLOGLOG, LOGIT, and PROBIT link functions are used for a logistic

regression analysis. The IDENTITY link function is used for a linear

regression analysis.

NOINT

Suppresses the intercept for the binary target model or the normal error linear regression model.

SINGULAR= n

Specifies the tolerance for testing singularity.

Default:

Copyright 2000 by SAS Institute Inc., Cary, NC, USA. All rights reserved.

---