Count outliers

* Question was: How can I show the number of outliers in a box plot (when many points are equal).
* Answer by rlevesque@videotron.ca.

INPUT PROGRAM.
SET SEED=98765413.
LOOP cnt=1 to 200.
COMPUTE var1=RV.NORMAL(100,10).
END CASE.
END LOOP.
END FILE.
END INPUT PROGRAM.

* Define 4 outliers having similar values.
IF $CASENUM<5 var1=150+$CASENUM/5.

* Round the values of var1 to the nearest 2.5.
* Rounding is necessary because if outliers are close to each other, the labels indicating 
* the number of values will overlap and will not be readable. Rounding will vary with
* the size of the variables; for large variables rounding might be to the nearest 100 or 1,000.
COMPUTE var1=RND(var1/2.5)*2.5.
AUTORECODE var1 /INTO var2.

SORT CASES BY var2 (A) .

DO IF $casenum =1.
	COMPUTE varlabel=1.
ELSE IF var2=lag(var2).
	COMPUTE varlabel=lag(varlabel)+1.
ELSE.
	COMPUTE varlabel=1.
END IF.

SORT CASES BY var2 (D) varlabel(D) .

DO IF $casenum =1.
	COMPUTE varlab=varlabel.
ELSE IF (var2<>lag(var2)) & (varlabel>1).
	COMPUTE varlab=varlabel.
END IF.
FORMATS varlab(F8.0).

EXAMINE
  VARIABLES=var1 /ID= varlab
  /PLOT BOXPLOT STEMLEAF
  /COMPARE GROUP
  /STATISTICS DESCRIPTIVES
  /CINTERVAL 95
  /MISSING LISTWISE
  /NOTOTAL.