Save regression equation for later use

[Ronein]

Hello

Please find example to  data set that check if the football  player scored goals in last game.

The dependent variable is GoalInd (binary) and there are 3 independent variables X1,X2,X3.

I used proc genmod to build regression model.

I want to use the regression equation that was created in order to calculate probability for new cases (data set new).

As I understand from my previous post there are 2 recommended way to do it:

Way1-Create a data set that store the regression equation  and then use it to calculate predicted values for new cases

Way2- Use store statement in proc reg and then use Proc PLM to  calculate predicted values for new cases

May you show please the code for way 1 and also the code for way 2?

thank you so much

 

Data Rawtbl;
input ID GoalInd X1 X2 $ X3 $;
cards;
1 0 20 W N
2 0 30 W N
3 1 20 B N
4 1 20 B N
5 0 30 B Y
6 0 20 B Y
7 1 20 W N
8 0 30 B Y
9 1 20 W N
10 1 30 B N
;
run;
/*Step1-Build regression equatiuon*/
proc genmod data=Rawtbl  namelen=60 descending;
class X1 X2  ;
model GoalInd=X1 X2  /dist=binomial link=logit  type3 wald;
output out=Build_NewModel1
p=P_BAD xbeta=logit;
run;

/*Step1-Test the model on a new data set*/
Data new;
input ID GoalInd X1 X2 $ X3 $;
cards;
11 0 30 B Y
12 0 20 W N
13 1 30 B N
14 1 20 B Y
15 0 30 B Y
;
run;
/*Way1 to save the regression equation and use it */
/*How can I save the Regression equation in a data set?*/
/*How can I use this data set to calculate predicted outcome for the new data set*/

/*Way2 to save the regression equation and use it */
/*How can I use store statement in proc genmod in order to save the resression equation?*/
/*How can I use these regression equation (that was stored in store statement in proc genmod) to 
calculate predicted outcome for the new data set*/

[Ksharp]

you can use option
proc genmod ...... outest=est
to save estimate coefficient and score the test data lately .

@Rick_SAS wrote many blog about this topic .

[Reeza]

One of Rick's blog posts that covers your question is here and outlines five different ways and some of the differences between them. 

https://blogs.sas.com/content/iml/2014/02/19/scoring-a-regression-model-in-sas.html

1. Missing response trick (ie combining regression + predictive data set)
2. PROC SCORE
3. SCORE statement within PROC
4. STORE Statement + PROC PLM
5. CODE statement - generates data step code to generate your predictive values. If you want to understand the math behind your formula this is a good optin. 

[Ronein]

Thank you, however when I add outest=est then I get an error

 
proc genmod data=Rawtbl outest=est namelen=60 descending  ;
class X1 X2  ;
model GoalInd=X1 X2  /dist=binomial link=logit  type3 wald;
output out=Build_NewModel1
p=P_BAD xbeta=logit;
run;

[Reeza]

There is no OUTEST option in PROC GENMOD. Please see the documentation.
https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=statug&docsetTarget=statu...

Other methods to do this have been shown by several other people so you'll need to use a different option.

[Ksharp]

Opps. My bad. I reckon that proc genmod like proc reg .

[PaigeMiller]

You can use the STORE command in PROC GENMOD or PROC GLM and then PROC PLM to get predicted values. Example: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.3&docsetId=statug&docsetTarget=statu...

[STAT_Kathleen · Accepted solution]

The following SAS Usage Note:

  http://support.sas.com/kb/33307

demonstrates 4 ways to score new observations using a previously fitted model .