topic Re: Missing value treatment with Regression estimates in Statistical Procedures

Missing value treatment with Regression estimates

KafeelBasha — Sat, 16 Jul 2016 11:11:09 GMT

Hello

While treating missing values I have came across Regression estimates to treat missing values.

proc reg data=data1 outset=data2;

repmis :model mv=v1 v2 v3 v4/selection=backward;run;

proc score data=data1 score=data2 out=sample type=parms predict;

var v1 v2 v3 v4;run;

We have replaced missing values by "repmis".

Can you please explain me above codes?. How proc reg helps us in finding the value of "repmis" and the need of proc score?.

Thanks

stat_sas — Sat, 16 Jul 2016 12:22:47 GMT

Hi,

Please provide some sample data. Also, expain your question a bit more.

Reeza — Sat, 16 Jul 2016 14:40:36 GMT

Repmis is the name of the model. MV is the variable being predicted.

The proc reg creates a model of MV based on the 4 variables included.

Proc score then calculates the predicted values for the variable MV based on the model from proc REG.

Rick_SAS — Sat, 16 Jul 2016 16:57:46 GMT

To understand the connection between PROC REG and PROC SCORE, see the article "Techniques for scoring a regression model in SAS." The label in the PROC REG step is used by PROC SCORE to name the predicted variable. You can find more examples in the documentation for PROC SCORE.