*-------------------------------------------------------------------*
| Purpose: To identify outliers based on n standard deviations      |
|          from the mean, computed from the middle 50% of the data. |                                             
| Arguments: DSN    - Data set name                                 |
|            VAR    - Numeric variable to be checked                |
|            IDVAR  - ID variable name                              |
|            N_SD   - The number of standard deviation units you    |
|                     would specify if the data values were not     |
|                     trimmed.                                      |
|                                                                   |
| EXAMPLE: %SD_TRIM(CLEAN.PATIENTS,HR,PATNO,2)                      |
*-------------------------------------------------------------------*;
%MACRO SD_TRIM(DSN,VAR,IDVAR,N_SD);

   TITLE1 "Outliers for Variable &VAR Data Set &DSN";
   TITLE2 "Based on &N_SD Standard Deviations Estimated from Trimmed  (50%)Data";

  PROC RANK DATA=&DSN OUT=TMP GROUPS=4;
      VAR &VAR;
      RANKS R;
   RUN;

   PROC MEANS DATA=TMP NOPRINT;
      WHERE R IN (1,2);  ***The middle 50%;
      VAR &VAR;
      OUTPUT OUT=MEANS(DROP=_TYPE_ _FREQ_)
             MEAN=M
             STD=S;
   RUN;

   DATA _NULL_;
      FILE PRINT;
      SET &DSN;
      IF _N_ = 1 THEN SET MEANS;
      IF &VAR LT M - &N_SD*S*2.65 AND &VAR NE .
         OR &VAR GT M + &N_SD*S*2.65 THEN PUT &IDVAR= &VAR=;  
   RUN;

   PROC DATASETS LIBRARY=WORK NOLIST;
      DELETE MEANS;
      DELETE TMP;
   RUN;
   QUIT;

%MEND SD_TRIM;