
	Where results make sense

Topic: Diehard tests

In the News (Wed 23 Dec 09)

Northern Trust

Marvin Harrison

Peter Chernin

Gary Locke

Match Play

Penelope Cruz

Mickey Rourke

AR Rahman

Danny Boyle

Hugh Jackman

	SPRNG: Scalable Parallel Pseudo-Random Number Generator Library
		The basic idea behind the statistical tests of randomness is that the random number streams obtained from a generator should have the properties of a sample drawn from a uniform distribution.
		DIEHARD tests implemented by Marsaglia are practically considered the standard tests.
		DIEHARD tests were proposed and implemented by Marsaglia, and are considered to be more stringent tests of randomness than the tests proposed by Knuth.
	sprng.cs.fsu.edu /Version2.0/statistical-tests.html (518 words)

	I recently tested the pseudo
		This test uses :: :: n=2^24 and m=2^9, so that the underlying distribution for j :: :: is taken to be Poisson with lambda=2^27/(2^26)=2.
		Then the quadratic form in the :: :: weak inverse of the 120x120 covariance matrix yields a test :: :: equivalent to the likelihood ratio test that the 120 cell :: :: counts came from the specified (asymptotically) normal dis- :: :: tribution with the specified 120x120 covariance matrix (with :: :: rank 99).
		Test numbers=0 mod 5 :: :: are printed but the KSTEST is based on the full set of 100 :: :: random choices of 8000 points in the 10000x10000 square.
	www.cse.csiro.au /poptools/diehard.htm (2829 words)

	Testing the PDH Random Numbers
		Several difference batteries of tests were carried out on the generators, including tests from SPRNG, the DIEHARD tests (as modified for SPRNG), the NIST tests, and certain empirical tests found in Knuth's chapter on random number testing.
		The conclusions from these tests specifically are that all the kk0 and SPRNG generators pass the tests while Random() fails two of the gap tests.
		However, in these test results we actually have performed many of each test, and the number reported for these tests is not just a raw score/probability, as above, but is the percentage of the tests that are passed in the group of each test.
	www.cs.fsu.edu /~mascagni/pdhtests.html (603 words)

	[No title]
		DIEHARD applies various methods of assembling and combining uniform random numbers, and then performs statistical tests that are expected to be nonsignificant; this suite of tests has become a standard method of evaluating the quality of uniform random number generator routines.
		This post hoc test (or multiple comparison test) can be used to determine the significant differences between group means in an analysis of variance setting.
		This post hoc test (or multiple comparison test) can be used to determine the significant differences between a single control group mean and the remaining treatment group means in an analysis of variance setting.
	www.statsoft.com /textbook/glosd.html (3132 words)

	Diehard battery of tests (Site not responding. Last check: 2007-10-11)
		Diehard tests are thereefore often reffered to as "stringent tests".
		Then the quadratic form in the weak inverse of the 120x120 covariance matrix yields a test equivalent to the likelihood ratio test that the 120 cell counts came from the specified (asymptotically) normal dis- tribution with the specified 120x120 covariance matrix (with rank 99).
		Testing is repeated using sequentially chunks of data from the file untill the file is exhausted.
	random.com.hr /products/random/manual/html/Diehard.html (2637 words)

	Top Secret Crypto Gold DieHard Test Results (Site not responding. Last check: 2007-10-11)
		An individual test can be considered passing if the p-value is between 0.025 and 0.975, forming a 95% confidence interval around the theoretical value specified within the test.
		However, to evaluate a data sample against the entire test suite requires consideration of all 250 p-values that are generated and the calculation of the probability that the entire suite passes with 95% confidence.
		Therefore, the RNG fails the Diehard tests if there is a p-value greater than or equal to 0.9999 or less than or equal to 0.0001.
	www.topsecretcrypto.com /DieHard.htm (1346 words)

	Hardware random number generators
		However, the most important is that the Diehard tests are designed for 32 bit words, whereas hardware random number generators tend to generate their bits one by one and don’t have a 32 bit structure.
		So the tests which are applied to the raw series, apart for the test for bias, must be insensitive to bias.
		A run length test, but not this one, is used in the FIPS 140-1 (section 4.11.1) test so I wanted a runs test in my set of tests.
	www.robertnz.net /hwrng.htm (4444 words)

	Robert G. Brown's General Tools Page
		Statistical Test Suite (STS) developed by the National Institute for Standards and Technology (NIST) are being incorporated, as are new tests developed by rgb.
		Note well that the primary objections I have towards diehard and sts are not that they are or are not adequate and complete; it is that the code itself is not properly packaged for reuse, testing, and extension.
		All tests are currently using the x.5.x test encapsulation and underlying KS and chisq tests.
	www.phy.duke.edu /~rgb/General/rand_rate.php (1405 words)

	Certification results
		The Stata pseudorandom-number generator was subjected to the Diehard suite of tests, which were developed by George Marsaglia and provide extensive testing of pseudorandom-number generators.
		Most of the tests in Diehard return a p-value, which should be uniform on [0,1) if the input file contains truly independent random bits.
		This specially formatted binary file was first produced, and then the Diehard tests were performed on the resulting file.
	www.stata.com /support/cert/diehard (488 words)

	Tests for Random Number Generators (Site not responding. Last check: 2007-10-11)
		The best way to test if a generator is good enough for an application is to run the application with two very different generators and see if they produce the same result.
		The sequence tested has to be at least as long as the sequence of random numbers your application plans to use, otherwise the biases your application may encounter can't be caught by the test.
		I developed these tests to break a generator, and I developed the generator to pass the tests.
	burtleburtle.net /bob/rand/testsfor.html (366 words)

	Tests for Randomness (Site not responding. Last check: 2007-10-11)
		Sadly, most of these tests are statistical in nature, and tend to concentrate on measuring bias.
		Note that a failure rate of 1/1000 means we should expect a good generator to "fail" if we test it 1000 times.
		The Only OK generator will pass most of the diehard tests, but a few of it's results have a less than.000000001 chance of occuring in an unbiased generator.
	www.helsbreth.org /random/tests.html (210 words)

	random.com.hr - Diehard
		The Diehard program written by B. Narasimhan can be found on the Web, but unfortunately seems not to have been maintained for the last few years.
		In hardware generators the main problems are the unavoidable bias (the probability of 1's is not the same as the probability of 0's) and bits correlation (a bit depends (statistically) on previous bits).
		Testing longer and longer sequences of random bits, the Diehard test will eventually detect these defects too, but specialized frequency and autocorrelation tests will do it faster thus saving the time.
	random.com.hr /products/random/Diehard.html (1898 words)

	True hardware random number generators
		diehard tests and estimating of simplifying details like an entropy be done.
		Because the diehard tests are most often stronger and more versatile than the other tests, they were used most often.
		diehard tests, without after-treatment and that this can be shown both experimental and in theory.
	www.schutzfehler.de (2483 words)

	RNGTS Home (Site not responding. Last check: 2007-10-11)
		Since the Ferrenberg affair it is known to a broad community that statistical tests alone do not suffice to determine the quality of a generator, but also application-based tests are needed.
		Most currently available test suites are limited to a subset of tests are written in Fortran or C and cannot easily be used with the C++ random number generator library.
		Test results are produced in an XML format, which through the use of XSLT transformations allows extraction of summaries or detailed reports, and conversion to HTML, PDF, PostScript or any other format.
	www.comp-phys.org:16080 /rngts (388 words)

	True random number generators
		Simulation of time-series that we are going to test for long-range dependence or chaos are examples where one might expect pseudo-random number generators to give misleading results.
		The testing on the final combinations of numbers simulating the draw from the urn would be aimed at just checking that the program was correct.
		It passed all the tests and I was unable to detect in deviations from perfect randomness.
	www.robertnz.net /true_rng.html (3868 words)

	[No title]
		To see how well such a simple and fast :: :: generator performs on tests of randomness, this program makes :: :: a large file with the multiply-with-carry generator implemen- :: :: ted in 16-bit integer arithmetic.
		It seems to pass all tests :: :: and is highly recommended for speed and simplicity.
		That means, :: :: of course, that the generator is likely to fail tests that :: :: depend, in any significant way, on the rightmost bit of a 32- :: :: bit random integer.
	www.isds.duke.edu /~rlw/marsaglia/makef.txt (1710 words)

	[No title]
		However, I could not find a definitive set of reported tests for the generator, so this morning, I wrote Java and C code to make simple tests, C code to generate data for the Marsaglia Diehard test suite, and C code to test the generator with the Marsaglia/Tsang Tuftest suite.
		There are failures in the OQSO (Overlapping Quadruples Sparse Occupancy) and DNA tests in the Diehard Battery suite, and in the Gorilla tests of the Tuftest suite; the failures there suggest poorer quality of 15 low-order bits (NB: both test suites number bits from high-order (leftmost) (0) to lowest (right-most) (31)).
		Applying the spectral test, it is possible to find bad subsequences with small step sizes for almost all linear pseudorandom number generators currently in use.
	www.math.utah.edu /~beebe/java/random/README (1316 words)

	Lucky skill
		Chi-squared test applied to a sample of die values obtained from the RNG indicated statistical randomness.
		Chi-squared test applied to a sample of pairs of die values (as used in backgammon) obtained from the RNG also indicated statistical randomness.
		While it is not possible to test all possible scenarios in a laboratory environment, iTech Labs has conducted a level of testing appropriate for a submission of this type.
	www.backgammonmasters.com /itechlabs/LuckySkill_RNG_report.htm (734 words)

	Re: NIST randomness-tests - EXE?
		I have embedded that test in a huge program to test the generators, so I can't send the executable, but if you send me the source code of the generator I can test it for you (in a proper way).
		Apart from the battery David Sexton > proposed me, I have rund the Diehard tests and now have uploaded the > results - at http://www.vmpcfunction.com/c7.htm Looking at those results one can only says: "Seem good...".
		To get a more useful result, you should run the test 50, 100 times and then you should see if the p-values gotten are evenly distributed.
	www.usenet.com /newsgroups/sci.crypt/msg03677.html (312 words)

	ArcView Random Numbers
		The DNA test indicates that bits 6 and 7 of each byte generated by Method 3 are the causes of the failure.
		The nature of the test failure is particularly awful: the frequencies of "1"s in bits 0-2 and 29-31 of the output were significantly different than the frequencies of "0"s, so the output is non-uniform!
		I have not tested this, in part because the algorithm used for RAND() (which Microsoft has published) is poor, so there's really not much hope.
	www.quantdec.com /arcview.htm (1906 words)

	Abstract: Isaacs_A.html
		Although many of these algorithms map well to our hardware, they either consistently fail statistical tests or are limited in their evolvable characteristics.
		Our off-line testing indicates that evolved generators of the ECRNG family routinely pass all tests in the Diehard battery.
		These are superior to many known, non-evolvable generators (for contrast, the C rand function routinely fails all 18 Diehard tests).
	klabs.org /richcontent/MAPLDCon02/abstracts/isaacs_a.html (596 words)

	SHAZAM DIEHARD test
		Below is the output of randomness tests for a sample of random numbers generated by SHAZAM.
		.975 means that the RNG has "failed the test at the.05 level".
		:: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: OPERM5 test for file random.bin For a sample of 1,000,000 consecutive 5-tuples, chisquare for 99 degrees of freedom= 76.482; p-value=.045287 OPERM5 test for file random.bin For a sample of 1,000,000 consecutive 5-tuples, chisquare for 99 degrees of freedom= 92.959; p-value=.347906 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :: This is the BINARY RANK TEST for 31x31 matrices.
	shazam.econ.ubc.ca /intro/diehard.htm (1811 words)

	brianbec's WebLog : Oh, the Brutality!
		The battery is called "DIEHARD" (nyuk nyuk), and encapsulates the accumulated experience and wisdom of George Marsaglia, a recognized expert.
		This is a crushing, brutal series of tests that looks for patterns in dozens of ways.
		For some reasons I could not diagnose, DIEHARD refused to run its last handful of tests on MY data sets, encountering some kind of end-of-file error, even though my files were of the same format (so far as I could tell) as Marsaglia's.
	weblogs.asp.net /brianbec/archive/2004/08/18/216818.aspx (1000 words)

	Re: AES and Diehard
		In Diehard all you need >> to do is to code an AES based PRNG for tesing.
		It's > entirely possible that a short seuquence of truly random bits fails a > test out of the law of probability, a bug in diehard or a test that is > flawed [e.g.
		I also doubt that Diehard tests will show up any issues in AES but I am happy to help others in their experiments.
	www.usenet.com /newsgroups/sci.crypt/msg01676.html (203 words)

	[No title]
		So keep in mind that "p happens" Enter the name of the file to be tested.
		Results of the OSUM test for B7-4_12M.1 Test no p-value 1 0.282338 2 0.094958 3 0.619320 4 0.005324 5 0.796188 6 0.459778 7 0.942950 8 0.021223 9 0.076473 10 0.768466 _____________________________________________________________ p-value for 10 kstests on 100 sums: 0.174324 ----------------------------------------------------------
		In particular, there may be some values so close to 0 or 1 that the tests they came from should be applied several more times, or new, related tests should be undertaken.
	qrbg.irb.hr /diehard2_qrbg121.txt (1570 words)

	[No title]
		From: scott@helsbreth.org (Scott Nelson) Subject: Re: obtaining proper input for DIEHARD rng tests Date: Thu, 06 Jan 2000 22:57:53 GMT Newsgroups: sci.math Summary: [missing] On 6 Jan 2000 peterw@ugcs.caltech.edu (Peter T. Wang) wrote: >Hi everyone, > >This is perhaps a bit off-topic, Nonsense.
		I've been told by a number of people that sci.math is the proper place to discuss the science of randomness and all that goes with it.
		Incidentally, I believe the "ArcView" generator discussed on that page--which passes all but one of the Diehard tests after it is suitably adjusted--may be derived from a generator built into Windows NT 4.0.
	www.math.niu.edu /~rusin/known-math/00_incoming/diehard (746 words)

	Description of the RAND function in Excel 2003
		The RAND function in earlier versions of Excel used a pseudo-random number generation algorithm whose performance on standard tests of randomness was not sufficient.
		Several of the Diehard tests produced unsatisfactory results with earlier versions of RAND because the cycle before numbers started repeating was unacceptably short.
		It failed several standard tests of randomness, making its performance an issue when a lengthy sequence of random numbers was needed.
	support.microsoft.com /default.aspx?scid=kb;en-us;828795&... (696 words)

	[No title] (Site not responding. Last check: 2007-10-11)
		::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :: TEST RESULTS FROM DIEHARD.EXE :: :: http://stat.fsu.edu/~geo/diehard.html :: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: NOTE: Most of the tests in DIEHARD return a p-value, which should be uniform on [0,1) if the input file contains truly independent random bits.
		= 8.55 p-value=.799293 ::::::::::::::::::::::::::::::::::::::::: The 9 p-values were.959512.562068.333483.002779.746810.938654.623554.758797.799293 A KSTEST for the 9 p-values yields.839674 $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :: THE OVERLAPPING 5-PERMUTATION TEST :: :: This is the OPERM5 test.
		:: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: OPERM5 test for file RANDOM.DAT For a sample of 1,000,000 consecutive 5-tuples, chisquare for 99 degrees of freedom=117.976; p-value=.906231 OPERM5 test for file RANDOM.DAT For a sample of 1,000,000 consecutive 5-tuples, chisquare for 99 degrees of freedom=120.491; p-value=.930004 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :: This is the BINARY RANK TEST for 31x31 matrices.
	www.sunny-beach.net /random_numbers/results.txt (1990 words)

	[No title]
		Java Randomness Test Suite is a gui application to run randomness tests on random stream resource (file or algorithm).
		Design is expandable and enable the user to add tests, input/output resources and algorithms.
		The adapters - tests, random streams, algorithms - are very simple interfaces, to be implemented by the user to enhance the randomness test suite.
	jrandtest.sourceforge.net (86 words)

Try your search on: Qwika (all wikis)