topic Re: Proc Logistic. Why is predicted probability different than calculated from a regression equation in SAS Programming

Proc Logistic. Why is predicted probability different than calculated from a regression equation?

Jedrek369 — Tue, 24 Aug 2021 16:24:51 GMT

As in a subject.

data have;
set sashelp.cars;
isUSA = Origin="USA";
run;

proc logistic data=have;
model isUSA=EngineSize Cylinders Horsepower;
output out=pred p=p;
run;

data want;
keep p p_manual;
set pred;
p_manual = 1/(1+EXP(-1 * 0.3199 -3.091*EngineSize 
+0.5735*Cylinders 
+0.0334*Horsepower));
run;

Why is p different than p_manual variable in the want dataset?

Thanks.

Re: Proc Logistic. Why is predicted probability different than calculated from a regression equation

Reeza — Tue, 24 Aug 2021 16:55:09 GMT

How different?
Here's an example showing results that match exactly.
https://communities.sas.com/t5/Statistical-Procedures/How-to-determine-logistic-regression-formula-from-estimates/td-p/120780

You were missing a set of brackets in your formula:

p_manual = 1/(1+EXP(-1 * (0.3199 -3.091*EngineSize 
+0.5735*Cylinders 
+0.0334*Horsepower)));

@Jedrek369 wrote:

As in a subject.

data have;
set sashelp.cars;
isUSA = Origin="USA";
run;

proc logistic data=have;
model isUSA=EngineSize Cylinders Horsepower;
output out=pred p=p;
run;

data want;
keep p p_manual;
set pred;
p_manual = 1/(1+EXP(-1 * 0.3199 -3.091*EngineSize 
+0.5735*Cylinders 
+0.0334*Horsepower));
run;

Why is p different than p_manual variable in the want dataset?

Thanks.

Re: Proc Logistic. Why is predicted probability different than calculated from a regression equation

Jedrek369 — Wed, 25 Aug 2021 07:33:51 GMT

Shoot. That's embarrassing. Thanks anyway.

Re: Proc Logistic. Why is predicted probability different than calculated from a regression equation

FreelanceReinh — Wed, 25 Aug 2021 12:03:12 GMT

Tip: The formula

p_manual = 1/(1+EXP(-1 * (...)));

can be simplified to

p_manual = logistic(...);

(see LOGISTIC function).