topic Re: Logistic Regression Forecasting in Statistical Procedures

Logistic Regression Forecasting

Crubal — Wed, 28 Jun 2017 21:29:35 GMT

Hi,

I have the following code using Logistic Regression on this dependent variable 'Stay', 1 means customer did stay; 0 means customer didnot stay. And several independent variables as below:

Proc Logistic Data=Work.Data1;
Class Member Travel_Agency;
Model Stay = Length_of_Stay Booking_Lead Member Travel_Agency Member*Travel_Agency / expb;
output out=Work.Data2 p=pi_hat;

Where Arrival_Date between '01MAY2014'd and '30APR2017'd;
Run;

My question is:

After it runs, in the output table, column 'pi_hat' is a value between 0 and 1. Does it mean it is more likely to be 'Stay' or 'Not Stay'? And how could I apply the result from logistic regression into future 'Arrival_Date'? Thanks!

Re: Logistic Regression Forecasting

Ksharp — Thu, 29 Jun 2017 13:14:21 GMT

'pi_hat' is Proability for Stay =0.

If you want Stay =1 proability. try

Model Stay(event='1') =

Re: Logistic Regression Forecasting

Crubal — Thu, 29 Jun 2017 13:31:02 GMT

Thanks Ksharp, I see.

And how could I apply it into future dates (from '01MAY2017' to '30APR2018') using 'pi_hat'?

Thank you!

Re: Logistic Regression Forecasting

Ksharp — Thu, 29 Jun 2017 13:46:47 GMT

I don't understand what you mean.

You mean SOCRE a new dataset ?

use SCORE statement.

model ........

score data=new_data .....