Scoring Wizard Applied to Cox Regression

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Scoring Wizard Applied to Cox Regression

Atai Winkler

Hi

 

I am having trouble reconciling the answers I obtain from the Scoring Wizard applied to the Cox regression model. To help me understand what is going on in the Wizard, I used the same set of variables in the /MAP and MODEL statements (see below). In this case I had expected the same answers (survival_prob and cum_hazard) from the Wizard as I got when I developed the model – after all I am using the same variables.

 

MODEL HANDLE NAME=survival_model

   FILE=!model_file_handle

  /OPTIONS MISSING=SUBSTITUTE

  /MAP VARIABLES=most_recent_intvn  no_ntes_since_mst_rcnt_te  no_prev_terminal_events_1  no_prev_terminal_events_2  no_prev_terminal_events_3

  MODELVARIABLES=most_recent_intvn  no_ntes_since_mst_rcnt_te  no_prev_terminal_events_1  no_prev_terminal_events_2  no_prev_terminal_events_3.

  COMPUTE survival_prob=APPLYMODEL(survival_model, 'PREDICT').

  COMPUTE cum_hazard=APPLYMODEL(survival_model, 'CUMHAZARD').

EXECUTE.

MODEL CLOSE NAME=survival_model.

Some of the values from the model and the Scoring Wizard are the same but I had expected ALL the answers to be the same.

 

So, why are only some of the answers the same? I can provide the model and a full dataset with explanation.

 

Thank you for your help.

 

Regards

 

Atai

 

 

Dr Atai Winkler

Principal Consultant
PAM Analytics

 

[hidden email]

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD