random_number

hi,i use ftn95 on windows 7,and the problem is that random_number does not give uniform distribution at all.the probabilty in the interval from 0 to 1 rises greatly any suggestions?

Does this example reproduce what you are finding ? integer4, parameter :: million = 1000000 real8 x, sx,sxx, n integer4 i ! sx = 0 ; sxx = 0 ! opps, programming error !!! do i = 1, million call random_number (x) sx = sx + x sxx = sxx + xx end do ! n = i-1.0d0 sx = sx / n sxx = sqrt ( (sxx - nsxsx) / (n-1) ) write (,1001)' Number of samples = ',n x = 0.5d0 write (,1002)' Average = ',sx, ' ( expected = ',x,' error = ',sx-x x = 1.0d0/sqrt(12.0d0) write (*,1002)' Standard deviation = ',sxx,' ( expected = ',x,' error = ',sxx-x 1001 format (a,f0.2) 1002 format (a,f0.9,a,f0.9,a,es14.6) ! end

    NO ERRORS  [<main program> FTN95/Win32 v4.9.0]
Creating executable: C:\\temp\\lgotemp@.exe
Program entered
 Number of samples  = 1000000.00
 Average            = 0.499785607 ( expected = 0.500000000 error =  -2.143933E-04
 Standard deviation = 0.288603871 ( expected = 0.288675135 error =  -7.126348E-05
      

The following code puts the first 1000000 random values in a histogram with 20 classes each of width 0.05. Combine this with John's code above which shows the sample mean and standard deviations match approximately the theoretical values and it would seem the numbers are reasonably uniform across 0 to 1.

My code is run on Windows Vista. I don't see how it would be different on Windows 7.

Can you post some code to show what you are doing?

PROGRAM ANON

   IMPLICIT NONE
   INTEGER, PARAMETER :: N = 1000000
   INTEGER I, BIN, HIST(20)
   REAL*8 X
   
   HIST = 0
   
   DO I=1, N
      ! GET RANDOM NUMBER X SUCH THAT 0.0D0 <= X < 1.0D0
      CALL RANDOM_NUMBER(X)
      ! Since X >=0 smallest possible BIN value is 1
      ! Since X < 1 largest possible BIN value is 20
      BIN = INT(X / 0.05D0) + 1
      HIST(BIN) = HIST(BIN) + 1
   END DO
   
   DO BIN=1, 20
      PRINT *, BIN, HIST(BIN)
   END DO
   
END PROGRAM

Resuts obtained are:

        1       50091
        2       49911
        3       49894
        4       50163
        5       49874
        6       50266
        7       50193
        8       49883
        9       50026
       10       49986
       11       50303
       12       49892
       13       50042
       14       49975
       15       49724
       16       49765
       17       49933
       18       50343
       19       50066
       20       49670

yes it is solved,thank you for attention but it was my mistake