topic Re: Using proc means output in datastep in New SAS User

Using proc means output in datastep

Axsta — Wed, 14 Apr 2021 03:33:43 GMT

I calculated the median using proc means and I would like to know how to use the median output from proc means in datastep to impute the missing values.
Using the output option creates a dataset and using set in the step is not what I intend. Is there anyway that single value can be used ?

Re: Using proc means output in datastep

japelin — Wed, 14 Apr 2021 03:57:17 GMT

Since proc means is a procedure for calculating basic statistics, I don't think it has the function to rewrite the original data set.

Re: Using proc means output in datastep

Tom — Wed, 14 Apr 2021 04:18:08 GMT

If you want to manipulate data having it in datasets is the way to go.

But if the goal is just to replace the missing values with the MEDIAN then just skip PROC MEANS and use PROC STDIZE with METHOD=MEDIAN and MISSING=REPLACE.

Re: Using proc means output in datastep

andreas_lds — Wed, 14 Apr 2021 04:51:11 GMT

If you don't want to use proc stdsize, merging the result of proc means with the original dataset could be an option.

Re: Using proc means output in datastep

Aku — Wed, 14 Apr 2021 05:21:35 GMT

With Proc SQL using self join and Coalesce function it's pretty easy to this.

Re: Using proc means output in datastep

Rick_SAS — Wed, 14 Apr 2021 13:35:17 GMT

For an example of using PROC STDIZE to impute by using the median, see the article "Mean imputation in SAS," which mentions the METHOD=MEDIAN option.

You can also get the value from PROC MEANS by reading it into a macro variable, but it requires more effort and more steps:

proc means data=sashelp.heart N NMISS MEDIAN stackods;
   var Cholesterol;
   ods output Summary=MedianOut;
run;

data _NULL_;
set MedianOut;
call symputx("Median", Median);
run;
 
%put &=Median;

data Impute;
set sashelp.heart;
if missing(Cholesterol) then 
   Cholesterol = &Median;
run;

proc means data=Impute N NMISS MEDIAN;
   var Cholesterol;
run;

Re: Using proc means output in datastep

ballardw — Wed, 14 Apr 2021 14:34:04 GMT

@Axsta wrote:
I calculated the median using proc means and I would like to know how to use the median output from proc means in datastep to impute the missing values.
Using the output option creates a dataset and using set in the step is not what I intend. Is there anyway that single value can be used ?

Since there at least two different methods to create a data set as output from Proc Means and they can have different structures depending on options chosen, the first thing you should do is show the Proc Means code used to generate the output data set.

Impute the missing values where? In the result of proc means? A different data set? Single value of what?

Re: Using proc means output in datastep

Axsta — Thu, 15 Apr 2021 11:03:32 GMT

I am preparing myself for the base sas programming specialist exam and this was one of the questions in them. I am not sure if they require you to use advanced macros and call symput.