Skip to main content

Table 3 Average runtime in seconds (s.e.) for the balanced experimental design in scenario 1 based on 100 replicates of the simulation

From: A new approach fits multivariate genomic prediction models efficiently

Methoda

Software

Modelc

Runtime

PEGS

RR

0.4 (0.0)

THGS

RR

0.3 (0.0)

UV-THGS

RR

0.2 (0.0)

AI-REML (EVD)

ASREML-R

GBLUP

3.3 (0.3)

AI-REML

ASREML 4.2

GBLUP

272.6 (36.5)

AI-REML

AIREMLF90

GBLUP

109.8 (2.4)

EM-REML

REMLF90

GBLUP

1250.7 (11.7)

Gibbs samplingb

GIBBS3F90

GBLUP

559.8 (9.6)

  1. aPEGS pseudo expectation Gauss–Seidel, THGS tilde-hat Gauss–Seidel, UV-THGS univariate-tilde-hat Gauss–Seidel, AI average information, REML restricted maximum likelihood, EVD eigenvalue decomposition, EM expectation maximization
  2. b10,000 MCMC iterations
  3. cRR ridge-regression, GBLUP genomic best linear unbiased prediction