Date: Sun, 22 Jan 2006 20:45:42 -0500 Phil Jackson "SAS(r) Discussion" Phil Jackson Re: Programming Help text/plain; charset=ISO-8859-1

On Sun, 22 Jan 2006 19:43:39 -0600, Jiann-Shiun Huang <Jiann- Shiun.Huang@AMERUS.COM> wrote:

>Phil: > > Let me see if I understand this correctly: > > First determine the highest probability. In this case, 0.96 in row >11 is the highest. After that, determine the parent of the node till >the root. > > Would you confirm or explain the process to get the desired path?

Exactly.

Thanks, Phil

> >J S Huang >1-515-557-3987 >fax 1-515-557-2422 > >>>> Phil Jackson <arlando20002000@YAHOO.COM> 1/22/2006 7:38:08 PM >>> >On Sun, 22 Jan 2006 19:29:14 -0600, Jiann-Shiun Huang <Jiann- >Shiun.Huang@AMERUS.COM> wrote: > >>Phil: >> >> If * is not in the output, then what is the criteria used to >>determine which row should be "starred", or selected? Do you use the >>product of all probabilities from the root to all leaves, or nodes >>without children, to determine the path? I.e., do you first >determine >>all the paths from the root to all leaves and then find the product >of >>all probabilities to determine the path with the highest probability? >>In this example, there are two paths: >>(1) root - 1 - 2 - 4 - 10 - 11 - 14 >>(2) root - 1 - 3 - 8 - 12 >> >> Would you confirm if the above is what you want? >> >The probabilities are not all that important initially. The only >probability that is important is the maximum probability. When I get >that >maximum probability, I will use the node and parent of that observation >and >work my way backward to find all observations in the parent-node >chain. >Only the highest probability is important, the important ones are the >parent and nodes. Your (2) is exactly what I need. > >Thanks, >Phil > > >>J S Huang >>1-515-557-3987 >>fax 1-515-557-2422 >> >>>>> Phil Jackson <arlando20002000@YAHOO.COM> 1/22/2006 7:18:07 PM >>> >>On Sun, 22 Jan 2006 22:57:12 +0100, Stéphane Colas >><saslist@DATAMETRIC.FR> >>wrote: >> >>>Phil, >>> >>>Could you test it ? >>> >>>proc sql noprint; >>>create table out_arbor as >>>select * >>>from arbor >>>where prob_target_10 like '%*' >>>order by node desc; >>>quit; >> >>The stars are not actually output, I only put the stars there for >>illustrations to show which rows I want. >> >>Thanks, >>Phil >> >>> >>>Stéphane. >>> >>>Selon Phil Jackson <arlando20002000@YAHOO.COM>: >>> >>>> I have a programming question that I need help with. I have some >>output >>>> from using the arbor procedure (decision tree). I want to get the >>path >>of >>>> greatest probability of getting a desired response. The output >>looks >>like >>>> this: >>>> >>>> node .. parent .. prob_target_10 ... >>>> 1 . .80* >>>> 2 1 .56 >>>> 4 2 .82 >>>> 3 1 .88* >>>> 8 3 .91* >>>> 10 4 .63 >>>> 11 10 .73 >>>> 12 8 .96* >>>> 14 11 .77 >>>> >>>> I need the path highlighted by stars and output those rows only. >>The >>node >>>> with the highest probability will always be the starting point, so >I >>have >>>> to work backwards to the starting node. They are linked by their >>node >>and >>>> parent identifications (12-8, 8-3, 3-1, 1-.). Any help would be >>greatly >>>> appreciated. >>>> >>>> Thanks, >>>> Phil >>>>

Back to: Top of message | Previous page | Main SAS-L page