New Contributor
Posts: 3

# Scoring Formula using Proc Logistic Output

I have received the following output in Proc Logistic:

 Variable DF Estimate StdErr WaldChiSq ProbChiSq StandardizedEst Intercept 1 -0.1744 0.2316 0.5671 0.4514 _ Var1 1 1.7537 0.2977 34.7055 <.0001 0.0772 Var2 1 0.0262 0.00393 44.3659 <.0001 0.0792 Var3 1 -0.4627 0.0773 35.8634 <.0001 -0.0813 Var4 1 -0.3209 0.049 42.8821 <.0001 -0.0884 Var5 1 -0.00191 0.000206 86.6722 <.0001 -0.1216 Var6 1 -0.7336 0.1117 43.1698 <.0001 -0.1282 Var7 1 -0.9459 0.1632 33.5811 <.0001 -0.057

How do I convert it into scoring formula?

I tried a formula based on the Estimate column:
-0.1744+1.7537*Var1+0.0262*Var2-0.4627*Var3 etc.. However, the result I am getting does not match the Score if I were to score using Proc Logistic (Proc Logistic Inmodel=a.Model)

I need to be able to create the score produced using Inmodel statement, but using linear function Intercept+XVar1+YVar2... Which info from the output I should be using and what is the transformation (if any) required?

Thank you!

Posts: 5,047

## Scoring Formula using Proc Logistic Output

What you get with your expression F = intercept + Var1*b1 + ... is Logit(p) = Ln(p/(1-p)). You want to estimate p = exp(F)/(1+exp(F)) = 1 / (1+exp(-F)).

PG

PG
Discussion stats