*Introduction to Categorical Data Analysis Procedures* |

## Logistic Regression

*Dichotomous Response*

You have many options for performing logistic regression
in the SAS System. For the dichotomous outcome, most
of the time you would use the LOGISTIC procedure
or the GENMOD procedure; you will need to code indicator
variables for classification effects in PROC LOGISTIC
but can use the CLASS statement in PROC GENMOD. The
LOGISTIC procedure provides model-building, so you may
choose to use it for that reason. (Note that a future
release of PROC LOGISTIC will include a CLASS statement).
You may want to consider the CATMOD procedure for logistic
regression since it handles classification variables;
however it isn't efficient for this purpose when you have
continuous variables with a large number of different values.
For a continuous variable with a very limited number of
values, PROC CATMOD may be useful. You list the continuous
variables in the DIRECT statement.

The PROBIT procedure also performs logistic regression, and
the LOGISTIC, GENMOD, and PROBIT procedures allow you to
use events/trials input for the responses; the ratio
of events to trials must be between 0 and 1.

*Ordinal Response*

The LOGISTIC and PROBIT procedures treat all response
variables with more than two levels as ordinal responses
and fit the proportional odds model. The GENMOD procedure
fits this model with a link function of CLOGIT and the
specification of the multinomial distribution.
*Nominal Response*

When the response variable is nominal, that is, there is
no concept of ordering of the values, you can fit a logistic
model to response functions called generalized logits.
Only the CATMOD procedure presently performs a generalized
logits analysis.

Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.