CART method in SAS

[pamelisa]

Dear Group,

i am attempting a classification tree for a medical outcome. both the outcome variable and independent variables are binary.

my tree looks like it is reading the program as if it is  a continuous independent variable (but it's actually 0|1).  For example, there are input values of 0.10 and 0.01.  i do not understand that, and the SAS example i read on line for classification trees show different outputs.

Does any one have some  idea of what could be going on here?  Also, was expecting a chi sq or p value. (?)

MY code:

data pam.inhib;
proc hpsplit data=pam.copyinhib;
class inhibition;
model inhibition =
acephate chlorpyrifos ddvp diazinon dimethoate methomyl mevinphos carbaryl carbofuran oxydemeton_methyl handler;
grow chaid;
run;

i've attached the tree

THANK YOU!

best, pam

[Reeza]

There's a lot of output there, can you specify the output that shows you have an issue, page and table/graph perhaps?

Specifically related to:

my tree looks like it is reading the program as if it is  a continuous independent variable (but it's actually 0|1).  For example, there are input values of 0.10 and 0.01.  i do not understand that, and the SAS example i read on line for classification trees show different outputs.

[StatDave]

All of the predictor variables are considered as continuous unless you also specify them in the CLASS statement. See the descriptions of the CLASS and MODEL statements in the PROC HPSPLIT documentation.

[pamelisa]

got it, great!!

thank you very much.

pam

[pamelisa]

Hi Reeza,

thanks for reaching back. The problem is fixed (it was the class statement).

i am new to this.  i see that SAS is splitting the tree only once now that i've fixed that statement.   my new question: is there a way to ask SAS to make more than one split?

best, pam