version date 1.01 8/1/95 Bug found in SIZE62. Incorrect when writing contrasts and not writing all solutions. Fix is to change line 771 from: SIZE62 = 0 to: SIZE62 = 2*NUMCNT 1.02 8/3/95 Bug found in write to MTGS61. Inverse elements of genetic uncorrelated random effect variance matrices were written rather than the matrix. Several changes were made. The basic change is to move the copy of the elements after the inversion back to the matrix when loading vector SAMPWT. 1.03 9/1/95 Minor revisions to manual. Printing of PS version done with Type 1 PS only. 1.04 9/12/95 Bug found with grouping random effects when user specifies blocks. The lines changes are marked with date (9/12/95) in subroutine GRPIND and BLDBLK in file MTGSRSUB.FOR. The bugs should only affect analyses where the user specified groups of uncorrelated random effects. 1.05 10/25/95 Bug found by Freddy Fikse in checking for valid range in random number generator seed. Note that the random number generator subroutine would catch the problem and stop the run if not caught by the Gibbs sampling program. Corrected lines are commented and marked with 10/25/95. 1.06 12/21/95 Make files added to distribution disk and compile batch files were changed. A set of utility programs have also been added to the distribution disk. See the UTILITY.DOC file for information about those programs. 1.07 2/21/96 Format of MTGS63 and MTGS65 changed to aid development of GIBANAL by Jan-Thijs van Kaam. GIBANAL will provide many tools necessary to determine burn-in, thinning, etc 1.08 5/16/96 Bug fixed in MTGSRUN related to shape parameters for residual variances. The incorrect line read: IF (SHAPEI(I).GT.0) THEN and the correct version is: IF (SHAPER(I).GT.0) THEN This should only cause problems in cases where non-flat priors were intended to be used for residual and flat priors were intended or non-existent for uncorrelated random effects (SHAPEI is the vector of shape parameters for uncorrelated random effects). This bug was discovered by Ismo Stranden. In addition, two new make files were included that are intended to be more portable. One is for PCs (makefile.pc) and one for UNIX (makefile.unx). 1.09 6/4/96 1) Bug fixed in MTGSPREP that causes an error in calculation of SD with different numbers of missing observations for multiple traits. The correction is to add the line: N = NRECY(I) in the DO loop for trait summary. At the current revision the line is inserted at line number 1473 following the line DO I = 1,NTRT An additional tool for finding the correct section of code is to search for 'Summary of data and mixed model'. This line precedes the DO loop by 13 lines. This bug was reported by Rainer Roehe. 2) A bug in the degrees of freedom for the inverted Wishart was corrected. The program incorrectly generated the IW variable from an IW distribution with 1 degree of belief too small. The bug is corrected by adding 1 to the degree of belief parameter when calling the WISHRT subroutine at lines 2715, 2773, and 3138-9 in the current version. The correct lines are (in order): CALL WISHRT(GSS,GIVEC,NANPSEC,NANIM+SHAPEG+1) CALL WISHRT(ISS(LL),RWRKVC1,MAPNUM(I),NRL(I)+SHAPEI(I)+1) CALL WISHRT(RSS(LL),RWRKVC1,RESNUM(I),RCNT(I)+SHAPER(I)+1) This bug was discovered by Ismo Stranden. Ismo suggested an alternative fix, which is to change the subroutine so that the subroutine generates Wishart variables corresponding to the DF rather DF + 1. 3) MTGSNRM was modified to use hashing rather than a simple lookup strategy. This makes the first pass MUCH faster for larger pedigrees. 1.10 7/3/96 Minor bug fixed in MTGSNRM. Hash vectors now initialized. 9/17/96 Minor changes in wording and information provided when a fixed or random effect level is not found in the binary search during the second read of the data. 4/14/97 Minor bug fixed in MTGSRUN when writing descriptive label out when system is restarted. Correction is at line 629 - change do loop variable to I. Bug found by Josef Siemes. 4/24/97 MSTIME subroutines changed for RS/6000 so that real times are reported. Modified subroutines provided by Mancang Dong. 1.11 5/23/97 Bug found and exterminated resulting from incorrect derivation of results when there are traits with missing values. 7/18/97 Removed CHISQU and CHLSLV (legacy code) from external commands due to compiler error noted by Chris Hagger. 9/7/01 Fixed problem in heritcor.for that was identified by Eldin Leighton. When phenotypic variances were added to the variables written to MTGS64, the information written to MTGS65 was not updated properly. 12/16/02 Fixed problem with MTGSNRM identified by Mike MacNeil. A variable used by the hash function was not initialized to zero, and can cause problems with some compilers. Problem was identified using g77.