WIND
      SUBROUTINE WIND(X,N,P1,P2,IWRITE,XWIND)
  
PURPOSE--This subroutine computes the sample windsorized mean of
         the data in the input vector X.  The windsorizing is
         such that the lower 100*P1 % of the data is replaced by
         the smallest non-windsorized value, and the upper
         100*P2 % of the data is windsorized.  Replaced by the
         largest non-windsorized value.
INPUT  ARGUMENTS--X      = The single precision vector of
                           (unsorted or sorted) observations.
                --N      = The integer number of observations
                           in the vector X.
                --P1     = The single precision value (between
                           0.0 and 1.0) which defines what
                           fraction of the lower order
                           statistics is to be windsorized
                           before computing the windsorized mean.
                --P2     = The single precision value (between
                           0.0 and 1.0) which defines what
                           fraction of the upper order
                           statistics is to be windsorized
                           before computing the windsorized mean.
                --IWRITE = An integer flag code which (if set to
                           0) will suppress the printing of the
                           sample windsorized mean as it is
                           computed; or (if set to some integer
                           value not equal to 0, like, say, 1)
                           will cause the printing of the sample
                           windsorized mean at the time it is
                           Acomputed.
OUTPUT ARGUMENTS--XWIND  = The single precision value of the
                           computed samplw windsorized mean
                           where 100*P1 % of the smallest and
                           100*P2 % of the largest ordered
                           observations have been windsorized
                           before computing the mean.
OUTPUT--The computed single precision value of the sample
        windsorized mean where 100*P1 % of the smallest and
        100*P2 % of the largest ordered observations have been
        windsorized.
PRINTING--None, unless IWRITE has been set to a non-zero integer,
          or unless an input argument error condition exists.
RESTRICTIONS--The maximum allowable value of N for this
              subroutine is 15000.
            --P1 should be non-negative.
            --P1 should be smaller than 1.0
            --P2 should be non-negative.
            --P2 should be smaller than 1.0
            --The sum of P1 and P2 should be smaller than 1.0.
OTHER DATAPAC   SUBROUTINES NEEDED--SORT.
FORTRAN LIBRARY SUBROUTINES NEEDED--None.
MODE OF INTERNAL OPERATIONS--Single precision.
LANGUAGE--ANSI FORTRAN.
REFERENCES--David, Order Statistics, 1970, Pages 126-130, 136.
          --Crow and Siddiqui, 'Robust Estimation of Location',
            Journal of the American Statistical Association,
            1967, Pages 357, 387.
          --Filliben, Simple and Robust Linear Estimation
            of the Location Parameter of a Symmetric
            Distribution (unpublished PH.D. Dissertation,
            Princeton University, 1969).
WRITTEN BY--James J. Filliben
            Statistical Engineering Laboratory (205.03)
            National Bureau of Standards
            Gaithersburg, MD  20899
            Phone--  301-921-2315
ORIGINAL VERSION--November  1975.
UPDATED         --February  1976.