topic Re: How to Use Regression to Fill missing Values like in SAS Programming

How to Use Regression to Fill missing Values like

saleh — Mon, 30 May 2016 19:28:46 GMT

How to Use Regression to Fill missing Values like this Examlpe

I have alot of missing values

but i don't want to replace it with static Value,as we can change the missing with mean and Standard Deviasion then i want to predict the Missing Value using an algorithm like Regression

simply i want to Merge First and Second code

data yourdata;

set yourdata;

array change _numeric_;

do over change;

if change=OldValue then change=newValue;

end;

run ;

Proc reg data=yourdata

Model=

Run

want to use the second Code replace the newValue

Re: How to Use Regression to Fill missing Values like

Reeza — Mon, 30 May 2016 18:45:42 GMT

I don't understand what your question is, please clarify. If you can post sample data and expected output.

Re: How to Use Regression to Fill missing Values like

PGStats — Mon, 30 May 2016 18:51:59 GMT

Don't change a missing value with zero unless zero truly is the observed value. A missing value signals the absence of information and is treated appropriatly.

Note: // is not the syntax for comments in SAS code.

Re: How to Use Regression to Fill missing Values like

saleh — Mon, 30 May 2016 19:13:28 GMT

I have alot of missing values

but i don't want to replace it with static Value,as we can change the missing with mean and Standard Deviasion then i want to predict the Missing Value using an algorithm like Regression

Re: How to Use Regression to Fill missing Values like

saleh — Mon, 30 May 2016 19:09:44 GMT

I would like to predict this missing value, How to do that ?

Thanks for your replan ....

Re: How to Use Regression to Fill missing Values like

Reeza — Mon, 30 May 2016 19:55:53 GMT

If you want to interpolate using regression you can use proc reg and get the output with the predicted value.

Here's an example using the CLASS dataset.

data class;
set sashelp.class;
if weight in (84, 112) then weight = .;
run;

proc reg data=class;
model weight = height age;
output out=want predicted=p;
run;quit;

proc print data=want;
run;