faq
Table of Contents
FAQ: examples, options, and tips for BLUPF90 family programs and modeling
Download BLUPF90 manual
Technical problems and segmentation fault
- Segmentation Fault: Received a message with “Segmentation fault”
- System Error: Received a message like “The program can't start because libiomp5md.dll is missing…” in Windows.
- Better settings for OpenMP: How should I define the number of threads used with a program?
- Installation of Intel Libraries: For users with customized programs
Frequent mistakes in data, pedigree, and parameter files for BLUPF90 programs
Data preparation program
REML
- blupf90+: This is a combination of remlf90, airemlf90, and blupf90
- remlf90: EM algorithm
- airemlf90: Average Information REML + REMLF90 including heterogeneous residual variances Standard Errors (PDF)
Gibbs Sampler
OPTIONs for priors and degrees of belief, for restarting from the last sampling, and for storing solutions and SD with or without fixed (known) variances
- gibbsf90+: This is a combination of Gibbs programs: gibbs2f90, gibbs3f90, thrgibbs1f90, and thrgibbs3f90
- gibbs1f90: Faster than GIBBS2F90 without missing traits
- gibbs2f90: Useful with missing traits
- gibbs3f90: With heterogeneous residual variances
- thrgibbs1f90: Threshold + linear model
- thrgibbs3f90: Threshold + linear model with heterogeneous residual variances
- General questions: Common to all Gibbs samplers
Post-Gibbs Analysis
- postgibbsf90: Statistics and graphics for convergence diagnosis and post-Gibbs analysis (use postgibbsold for old Gibbs sampler programs). The graphs are created with gnuplot, which must be pre-installed.
BLUP
- blupf90+: This is a combination of remlf90, airemlf90, and blupf90
- blup90iod3 (new blup90iod) - a combined program of blup90iod2, blu90iod2r, blup90iod2hr, and blup90mbe2 with new features
- cblup90iod2: (G)BLUP with threshold + linear model for a large number of equations (available only under some conditions)
- predictf90: Calculates y-hat
- blupf90: (G)BLUP with SOR, FSPAK or PCG
- blup90iod2: (G)BLUP for a large number of equations (available only under some conditions)
- blup90iod2r: (G)BLUP with GIBBS3F90-type heterogeneous residual variance for a large number of equations (available only under some conditions)
- blup90iod2hr: (G)BLUP with AIREMLF90-type heterogeneous residual variance for a large number of equations (available only under some conditions)
- blup90mbe2: (G)BLUP for a large number of equations for multi-breed evaluations (available only under some conditions). The prior external EBV can be from other breeds, lines, foreign origins, …
Accuracy (approximation)
- accf90: Accuracy (reliability) for a large number of animals (available only under some conditions)
- accf90GS: Accuracy (reliability) for a large number of animals with genotypes (available only under some conditions); requires options
- accf90GS2: New accuracy (reliability) for a large number of animals with genotypes (available only under some conditions); requires options
Genomic Analysis
Additional programs
- inbupgf90 - a program to calculate inbreeding coefficients with incomplete pedigree
- seekparentf90 - a program to verify paternity and parent discovery using SNP markers
- validationf90 - a program to validate models
- probcatf90 - a program to transform solutions from linear or threshold model to probabilities
Modeling
* Sire model (cg: contemporary group) y = μ + cg + ... + sire + e y = μ + cg + ... + sire + MGS + e where cg = common environment (farm, year, season, management, ...) * Animal model (pe: permanent environment) * Single record / animal y = μ + cg + ... + animal + e * Multiple records / animal y = μ + cg + ... + animal + pe + e * Random regressions f(t) y = μ + cg + ... + f(t1) + f(t2)*animal + f(t3)*pe + e where f(t) = a function of t (t: continuous variable) * Maternal effect model (mpe: maternal permanent environment) y = μ + cg + ... + animal + maternal + mpe + e
faq.txt · Last modified: 2024/10/23 21:56 by shogo