How to calculate decision weights in Decision node SAS Miner

msf2021 — Thu, 16 Sep 2021 09:58:29 GMT

Hi everyone,

I am learning how to use oversampling methods in SAS Miner and I am using this tutorial : https://documentation.sas.com/doc/en/emref/14.3/p0ak10c0ywr8mrn1an7tum03sp7g.htm#p1exm858amo9ren1k17yz8e0qu5s

I completly understant when to use the decisions node and what the priors are but when they explain how to fill the decision weights, i don't understand how they calculate the inverse prior:

"enter the decision weights as shown in the figure above. For level GOOD, column DECISION1, enter the inverse prior value for the original proportion (0.05) of rare events: 1.0526. For level BAD, column DECISION2, enter the inverse value prior for the original proportion (0.95) of non-events: 20.0. All other values in the matrix should be 0.0. Click OK. Now your Decisions node is configured."

Can someone help me here?

Re: How to calculate decision weights in Decision node SAS Miner

sbxkoenk — Thu, 16 Sep 2021 14:39:41 GMT

Hello,

Using inverse priors in a profit matrix yields the central cutoff: cutoff = prior1.

The central cutoff maximizes the sum of SE and SP -- a good thing, and a good choice.

SE = Sensitivity

SP = Specificity

Kind regards,

Koen

topic How to calculate decision weights in Decision node SAS Miner in Statistical Procedures

How to calculate decision weights in Decision node SAS Miner

Re: How to calculate decision weights in Decision node SAS Miner