MacAnova is conceived and programmed by Gary W. Oehlert and Christopher Bingham, Department of Applied Statistics, University of Minnesota, and is Copyright (C) 1994 - 2001 by them. Their e-mail addresses are kb@stat.umn.edu and gary@stat.umn.edu. MacAnova is distributed under the terms of the GNU Public License, Version 2 (see file COPYING distributed with MacAnova). Briefly, this means that MacAnova may be freely copied and distributed and the source is available. The source will be available by anonymous ftp or other equivalent retrieval from stat.umn.edu. Any changes others make to the source must be clearly marked. There is no warranty of any kind for MacAnova, either expressed or implied. MacAnova is distributed "as is". See file COPYING for a more complete statement. The MacAnova WWW home page is http://www.stat.umn.edu/macanova/macanova.home.html Executable versions of the Macintosh, DOS and Windows versions are available there, along with source and Portable Document Format (PDF) versions of the User's Guide and other documentation. An up-to-date mirror of these files is maintained by statlib at http://lib.stat.cmu.edu/ Reports of bugs should be emailed to kb@stat.umn.edu. The Macintosh version uses TransSkel 3.12, a transportable Macintosh application skeleton placed in the public domain by Paul Dubois (dubois@primate.wisc.edu). The extended memory MSDOS version (DJGPP) is compiled using a version of Gnu gcc developed and copyrighted by D. J. Delorie (DJGPP) and distributed under the terms of the GNU Public License. Starting with MacAnova 4.04, version 2 of this compiler has been used. Source and executable for DJGPP can be retrieved via anonymous ftp from oak.oakland.edu (in pub/simtelnet/gnu/djgpp); modifications to the DJGPP library for its use in MacAnova here can be retrieved via anonymous ftp from umnstat.stat.umn.edu. The Windows/Motif versions make use of the wxWin cross-platform windowing interface developed Dr. Julian Smart, Artificial Intelligence Applications Institute, The University of Edinburgh. wxWin is Copyright (c) 1995 Artificial Intelligence Applications Institute. The WxWin home page is http://web.ukonline.co.uk/julian.smart/wxwin/ . Currently both the Windows and Motif versions are based on version 1.68 of wxWin, a copy of which is available through the MacAnova home page. Plotting is done using a modification of GNUplot, Copyright (C) 1986, 1987 Thomas Williams, Colin Kelley. The Unix/Linux version and the extended memory DOS version (DJGPP) allow command line editing and history maintenance using the GNU Readline Library, Copyright (C) 1988, 1991 Free Software Foundation, Inc., distributed under the terms of the GNU public license. A compressed tar archive of version 2.0 (used in the Unix/Linux version) is available through the MacAnova home page. The version used in the DOS DJGPP version was included with the source for gdb4.12 found on ftp://oak.oakland.edu/ which has been reorganized since we retrieved it. Included in MacAnova's distribution are modified translations from Fortran to C of the following programs written by others. Program screen and related subroutines for computing regressions by leaps and bounds by G.M.Furnival and R.W.Wilson supplied by Sanford Weisberg. See their paper, Regression by Leaps and Bounds, Techno- metrics 16 (1974) 499-511. Subroutines rebak, reduc, rsg, tql2, tqlrat, tred1, tred2, svd, tridib, and tinvit from the Eispack library. Subroutines dchdc, dgeco, dgedi, dgefa, dgesl, and dqrdc from the Linpack library. Subroutines for computing mixed radix fast Fourier transforms written by Gordon Sande at the University of Chicago circa 1968. Program hc and related subroutines for computing hierarchical cluster analysis by F. Murtagh, retrieved from statlib. Subroutines for making stem and leaf displays from the book ABCs of EDA by David Hoaglin and Paul Velleman, Duxbury 1981. Subroutines to compute the roots of real polynomials from Algorithm 493 published in TOMS retrieved from netlib. Code to compute the cumulative normal adapted from W. J. Kennedy and J. E. Gentle, Statistical Computing, Marcel Dekker, 1980, pp 90-92, which is based on W. J. Cody, Rational Chebyshev approximations for the error function, Math. Comp 23 (1969) 631-637. Code to compute the inverse of a normal distribution from Algorithm AS 111 by J.D. Beasley and S. G. Springer, Appl. Statist. 26 (1977), 118-121 retrieved from statlib. Code to compute the inverse Student's t-distribution from CACM Algorithm 396, by G. W. Hill retrieved from netlib. Code to compute the (central) Beta distribution from a subroutine of W. Fullerton, Los Alamos, based on Bosten and Battiste, Remark on Algorithm 179, CACM 17 (1974) p. 153 Code to compute the inverse Beta distribution from Algorithm AS 109 by G. W. Cran, K. J. Martin and G. E. Thomas, Appl. Statist. 26 (1977), 111-114 retrieved from statlib. Code to compute the non-central Beta distribution from Algorithm AS 226 by R. V. Lenth, Appl. Statist. 36 (1987) 241-244, incorporating changes by H. Frick, Appl. Statist. 39 (1990) 311-12, retrieved from statlib Code to compute the gamma and chi-squared cumulative distributions from Algorithm AS 91 by D. J. Best and D. E. Roberts, Appl. Statist. 24 (1975), 385-388, incorporating revisions by B. L. Shea, Appl. Statist. 40 (1991), 233-235), retrieved from statlib. Code to compute the non-central chi-squared cumulative distribution from Algorithm AS 275 by Cherng G. Ding, Appl. Statist. 24 (1992), 478-482, retrieved from statlib. Code to compute the non-central Student's t cumulative distribution from Algorithm AS 243 by Russell V. Lenth, Appl. Statist. 38 (1989), 185-189, retrieved from statlib. Code to compute the cumulative distribution function and its inverse for the Studentized range from Algorithm AS 190 by R. E. Lund and J. R. Lund, Appl. Statist. 32 (1983) 204-210, incorporating corrections by Lund and Lund, Appl. Statist. 34 (1985) 104 and I. D. Hill, Appl. Statist. 36 (1987) 119, retrieved from statlib. Code for a combined uniform pseudo-random number generator for 32 bit machines in P. L'Ecuyer 1988 Comm. ACM, retrieved from netlib. Code implementing the Singleton quicksort algorithm (Comm. ACM Algorithm 347) adapted from ssort.f in cmlib. Code computing the cumulative distribution for Dunnett's t was adapted from Algorithm AS 251 by C. W. Dunnett, Appl. Statist. 38 (1989) 564-579 incorporating a correction by C. W. Dunnett, Appl. Statist. 42 (1993) p. 709, and subroutine mvstud, also by Dunnett, that is part of the AS 251 distribution from statlib. Code generating a pseudo-random Poisson variable adapted from a Fortran program in C. D. Kemp and W. A. Kemp, Poisson random variate generation, Appl. Statist. 40 (1991) 143-158. Code generating a pseudo-random binomial variable adapted from Algorithm 678, Transactions on Math. Software 15, 394-397 by Voratas Kachitvichyanukul and Bruce Schmeiser. Code implementing varimax rotation from subroutine varmx supplied by Douglas Hawkins (doug@stat.umn.edu). Code implementing k-means clustering from subroutine trwcla supplied by Douglas Hawkins (doug@stat.umn.edu). Code used to compute inverses to cumulative distributions from subroutine fsolve supplied by Douglas Hawkins (doug@stat.umn.edu). It is used by invchi() to compute the inverse of non-central chi-squared and by invdunnett() to compute probability points of Dunnett's t. Certain macros are also based on Fortran code: Macro levmar in file Arima.mac is based on a Fortran program of Ken Brown. See Brown,K.,M. and Dennis,J.,E., Derivative free analogues of the Levenberg-Marquardt and Gauss algorithms for nonlinear least squares approximation. Numerische Mathematik, Vol. 18, pp. 289-297 (1972) Macro neldermead in file Math.mac is based on Fortran subroutine MINIM by D. E. Shaw, CSIRO, Division of Mathematics & Statistics, with amendments by R. W. M. Wedderburn, Rothamsted Experimental Station, and Alan Miller, CSIRO, Division of Mathematics & Statistics. See also Nelder & Mead, The Computer Journal 7 (1965), 308-313. MINIM was retrieved from statlib.

Gary Oehlert 2003-01-15