Visualizing Categorical Data: interact
$Version: 1.0 (18 Mar 2001)
Michael Friendly
York University
Create interaction variables
The INTERACT macro creates interaction variables, formed as the product of
each of the variables given in one set (V1=) with each of the variables given in a second set (V2=).
The INTERACT macro is called with keyword parameters. The arguments may be
listed within parentheses in any order, separated by commas. For example:
%interact(v1=age sex, v2=I1 I2 I3);
- DATA=
-
The name of the input dataset. If not specified, the most recently created
dataset is used.
- V1=
-
Specifies the
name(s)
of the first set of
variable(s).
- V2=
-
Specifies the
name(s)
of the second set of
variable(s).
- OUT=
-
The name of the output dataset. If not specified, the new variables are
appended to the input dataset.
- PREFIX=
-
Prefix(s)
used to create the names of interaction variables.
The default is 'I_'. The names are of the form I_11 I_12 ...
I_1m I_21 I_22 ... I_nm
, where there are n variables in V1 and m variables in V2.
- CENTER=
-
If non-blank, the V1 and V2 variables are mean-centered prior to forming
their interaction products.
Example
Here we generate dummy variables for GROUP (GA and GB) and SEX (SF), and their interactions, SG11 and SG12.
%include vcd(interact); *-- or include in an autocall library;
%include vcd(dummy);
data test;
input y group $ sex $ @@;
cards;
10 A M 12 A F 13 A M 18 B M 19 B M 16 C F 21 C M 19 C F 20 C M
;
%dummy (data=test, var=sex group, prefix=s g) ;
%interact(data=test, v1=sf, v2=ga gb, prefix=sg);
proc print; run;
This gives
OBS Y GROUP SEX SF GA GB SG11 SG12
1 10 A M 0 1 0 0 0
2 12 A F 1 1 0 1 0
3 13 A M 0 1 0 0 0
4 18 B M 0 0 1 0 0
5 19 B M 0 0 1 0 0
6 16 C F 1 0 0 0 0
7 21 C M 0 0 0 0 0
8 19 C F 1 0 0 0 0
9 20 C M 0 0 0 0 0
See also
dummy Macro to create dummy variables
Previous: inflogis Next: label Up: Appendix A Top: index.html