BNET Procedure

As you review the network, you notice that Pain is the parent of Age. Given the data, there is a strong relationship between Age and Pain that explains this connection, but the direction is not as desired. You can use the INNETWORK= option as in the following statements to guide the network so that both Treatment and Age become parents of Pain:

Data Innetwork;
    length parent $ 10;
    input Parent $ Child $ Flag;
datalines;
Treatment  Pain   1
Age        Pain   1

data mylib.innetwork;
   set innetwork;
run;

 proc bnet data=mylib.neuralgia numbin=3 outnetwork=mylib.network
     innetwork=mylib.innetwork;
     target Pain;
     input Treatment Sex/level=nominal;
     input Age Duration/level=interval;
     ods output varselect=varselect;
 run;

 proc print data=mylib.network noobs label;
     var _parentnode_ _childnode_;
     where _type_="STRUCTURE";
 run;

The revised network is as desired, with Treatment and Age as parents of Pain, as shown in Figure 4.

You can get the revised conditional probabilities associated with this network by running the following code:

 proc print data=mylib.network noobs label;
     var _parentnode_ _parentcond_ _childnode_ _childcond_ _value_;
     where _type_="PROBABILITY";
 run;

Figure 5 shows the revised conditional probabilities, given this network.