You can create a project
input table either from the train table that you used to develop your
model, or you can define the project variables in a DATA step. The
project input table must include the input variables that are used
by the champion model. Therefore, if you have several candidate models
for your project, make sure that all candidate model input variables
are included in the project input table. If you create the project
input table from the train table, be sure to exclude the target variable
from the project input table.
Here is one method that
you can use to create the project input table from the train table.
Use the SET statement to specify the train table and the DROP or KEEP
statements to specify the variables from the train table that you
want in the project input table. You can drop the target variable
or keep all variables except the target variable.
This DATA step creates
the project input table from the train table and drops the target
variable Bad:
data hmeqtabl.invars;
set hmeqtabl.training (obs=1);
drop bad;
run;
This DATA step creates
the project input table from the train table and keeps all variables
except for the target variable Bad:
data hmeqtabl.invars;
set hmeqtabl.training (obs=1);
keep mortdue reason delinq debinc yoj value ninq job clno derog clag loan;
run;
You can also create
the project input table using the LENGTH statement to specify the
variables and their type and length. You could also specify the LABEL,
FORMAT, or INFORMAT statements, or the ATTRIB statement to specify
additional variable attributes. The following DATA step uses the LENGTH
statement to specify the project input variables in the table:
data hmeqtabl.invars;
length mortdue 8 reason $7 delinq 8
debinc 8 yoj 8 value 8
ninq 8 job $7 clno 8 derog 8
clag 8 loan 8;
run;
If you find that you
need to modify the project input variables after you have created
a project input table, you can use the project’s
Variables page
to modify the project variables.
For more
information, see
Defining Project Input and Output Variables.