Modification Records


March 12, 2004       Adjustment of histogram scales in 'rndkink' and 'rndkink2'.

March 13, 2004       Change line search of alpha's from single to triple to widen the parameter range
                     of log-normal tree: 28_5c,28_6a,28_6b,28_6c,ex28_10a,ex28_10b in Hull.

March 16, 2004       Some addition of comment and another way: gam.txt, erlang.txt, chisq.txt in Mun.

March 21, 2004       Addition of direct simulation method to hypergeo.txt in Mun.

March 22, 2004       Addition of direct simulation method to poisson.txt in Mun.
                     Modification of norm.txt by moment-match technique to norm_mm.txt in Mun.

March 26, 2004       Bug fixed and add mid-point stratified sampling to distributions 5B:
                     uniform.txt, norm.txt through triang.txt all in Mun.

Sept   8, 2004       Change in parameter setting in expo.txt with mean from 1/lambda to lambda in Mun.

Sept  10, 2004       Method change on triang(a,b,c) in accordance with CLT in Mun.

Sept  11, 2004       Same change on trigen(a,b,c) in Mun.

Sept  12, 2004       Bug fixed to alpha-percentile options(bbalpha.txt,mcalpha.txt,and bcalpha.txt) 
                     in Distinguished Parts, and some addition.

Sept  19, 2004       Parameter change on weibull.txt in Mun and related sections in Vose.

Sept  24, 2004       Adjustment of total number of random variables in custom.txt, customg.txt,
                     and rndkink.txt, i.e add just 'x=x[1:nn];' in all composition methods.

Sept  25, 2004       Add another version of geometric, geo2.txt, in Mun, and related adjustment in Vose.

                     Replace combination nCr to nCs by taking a log not to blow up.
                     5_2_1a.txt, bino.txt, hypergeo,txt and others in Vose.

Oct   20, 2004       trigen.txt, dtrigen.txt and tri1090.txt in Vose were completely wrong. Relplace f(x).

                     Adjustment of dtriang.txt in Vose to the cases with a=b and b=c as we didn't calculate.

Nov   29, 2004       All programs related to triang2 and triang2mp had a warning of two i's in the local 
                     statement, which was not detected by earlier versions of GAUSS I use. Now, one i is 
                     removed from each program in Vose and Mun. You might face some similar warnings. 
                     You could either ignore or remove one out of two locals.

Dec   29, 2004       All programs related to procedure phi(x) are modified to enable us to calculate one
                     column vector data in Ross: 2.2v, 2.2vi, 2.3a, 2.3b, 2.3c, 2.3d, 2.4a, 2.4i.

Feb       2005       Feb Maintenance: Typos in Distinguished Parts. Change all quasi program in Minotani
                     to Haltonb based(allow nn=10000). Erase american types from Distinguished Parts.

Mar       2005       March Maintenance: Faure sequence program is modified to Kijima(1996) based.
                     Both bisampler and trisampler in Distinguished Parts are modified in algorithm.
                     Now both 'beuro.txt' and 'teuro.txt' in deterministic binomial and trinomial will not 
                     blow up even when nn is large. Recursive vector index applied.

Apr       2005       April Maintenance: Both Fibonacci 1D and 2D in Core Parts are fixed. Lack of the last
                     element for both. Now, their means are 0.5 sharp. Begin out public on programs in
                     Japanese text. Copyright statement and vision are revised.

Apr   23, 2005       Scrambled programs in Distinguished Parts are fixed. Related programs:(siLHSHalton,
                     siLHSFaure, siLHSSobol, sHaltoninLHS, sFaureinLHS, and sSobolinLHS ) are fixed.

May    3, 2005       Use s=sqrt(sumc((x-xbar)^2)/n); where it should be divided by n instead of (n-1) in
                     JB.txt and CJB.txt(Jarque Bera Tests) in Classical Tests. We need that s to replicate
                     functions in most software.

Oct    7, 2005       Some rearrangement on Core Parts of Simulation. Change 2D LHS and 2D mid-point LHS 
                     from the versions of 1D twice to dim=2 cases of multi-dimensional LHS. Change the
                     examples of Twister, Swapper, and Scrambler from LHS to Hammersley to emphasize
                     that these add some randomness to deterministic numbers. Add Hypercube Vibrator.

                     Add '?' help statements including some notice to warn recent copyright infringements.

Mar   11, 2006       Important correction: 
                     Emprical Monte Carlo using a liner interpolation is completely WRONG in a theoretical
                     sense because we cannot convert the ordinal statistic to cardinal one. That's why this
                     method is so simple but not implemented. Especiallly, it is obvious if we consider the
                     updated "cdf". There are several papers and a book in Japan and a software in US dealing
                     with some variant of this, but all should be dropped. Especially, they should be dropped 
                     academically. That's partially my responsibility though I do not argue which one is first
                     to come. Academically worng is nothing but wrong. All we can do is to blur the realized 
                     ordinal points(Option 0). 
                     All the programs in distinguished part are rewritten and changed to the blur version.

May   18, 2006       Both tema and zlema in the section of smoothing are fixed to their well-known versions.