S******y 发帖数: 1123 | 1 I have a data set A with binary reponse variable 'event' - 0, 1 (i.e.
failure, success).
I already have prior knowledge of effects from X1 and X2 as baseline. i.e.,
I know -
log (p/(1-p)) = 0.34 + 0.25 * X1 + 0.81 * X2
The parameters above were obtained from previous analysis on a different
data set B.
Now I would like to evaluate the lift I can gain from including additional
variable X3. So here is what I did -
data A;
set A;
z = 0.34 + 0.25 * X1 + 0.81 * X2;
run;
proc logistic data=A;
model | D******n 发帖数: 2836 | 2 i guess for model 2 u should dump all the variables to fit a new model, rest
ricting the coefficients is weird,
.,
【在 S******y 的大作中提到】 : I have a data set A with binary reponse variable 'event' - 0, 1 (i.e. : failure, success). : I already have prior knowledge of effects from X1 and X2 as baseline. i.e., : I know - : log (p/(1-p)) = 0.34 + 0.25 * X1 + 0.81 * X2 : The parameters above were obtained from previous analysis on a different : data set B. : Now I would like to evaluate the lift I can gain from including additional : variable X3. So here is what I did - : data A;
| j*****e 发帖数: 182 | 3 Try this:
proc logistic data=A;
model event = X3/offset=z;
run; |
|