Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
Luis A. Afonso
Posts:
4,276
From:
LIsbon (Portugal)
Registered:
2/16/05
|
|
Checking a RNG by H (Fisher)
Posted:
Dec 4, 2012 8:56 AM
|
|
Checking a RNG by H (Fisher)
Fisher propose to use to check the p-values issued from NHST multiple tests through the r.v. H because the quantity H (Fisher) = -2*Sum (1, n) LOG (U) follows a 2*n degrees of freedom Chi-square distribution.
For each attempt 400´000 samples size 40 of RND´s were obtained and H calculated.
The interval [57.1532, 106.6286] should contain exactly 95% of H´s (Chi-square 80 d.f.).
Results:
__attempt 1, 2, 3, 4 : 0.94989, 0.95014, 0.95012, 0.94979.
__mean, standard error : 0.9500 , 0.0001
Luis A. Afonso
REM "HH7"
CLS
DEFDBL A-Z
RANDOMIZE TIMER
PRINT "____________HH7______________"
PRINT " AIM : Check Generator by samples size N ";
PRINT " after --> H(Fisher)=SUM(1,N).-2*LOG(RND) = CHI2";
PRINT " must [QUI(.025), QUI(.975)], 2N d.f."
PRINT " contains inside 95% "
REM
PRINT
INPUT " size=40 "; n
INPUT " repeat= "; ali
REM
INPUT " left QUI(.05) 57.1532"; left
INPUT " right (.95) 106.6286"; right
DIM x(n)
RANDOMIZE TIMER
FOR t = 1 TO ali
REM
LOCATE 15, 50: PRINT USING "########"; ali - t
H = 0
FOR j = 1 TO n
3 x(j) = RND
IF x(j) > 1 - 1E-15 THEN GOTO 3
LOCATE 15, 2
H = H - 2 * LOG(1 - x(j))
NEXT j
IF H > left AND H < right THEN inn = inn + 1
NEXT t
PRINT inn / ali
END
|
|
|
|
