A modification of REMLF90 for estimating variances with the Average-Information algorithm. Initially written by Shogo Tsuruta in 03/99-07/99. AIREMLF90 uses a second derivative REML algorithm with extra heuristics, as is described in Jensen et al. (1996-7). For most models, it converges in far fewer rounds than EM-REML as implemented in REMLF90. While typically REMLF90 takes 50-300 rounds to converge, AIREMLF90 converges in 5-15 rounds and to a higher accuracy. The final results will be saved in “airemlf90.log”.
See PREGSF90 with genotypes (SNP) for options.
OPTION conv_crit 1d-10
Convergence criterion (default 1d-12).
OPTION maxrounds n
Maximum rounds (default 5000). When n = 0, the program calculates BLUP without iterating REML and provides some statistics (-2logL, AIC, SE for (co)variances, …).
OPTION EM-REML n
Run EM-REML (REMLF90) for first n rounds to get initial starting variances for AIREMLF90 within the parameter space (default 0). With n is large (e.g., 1000, 10000, ….), AIREMLF90 runs as REMLF90 until convergence, and then switching back to AIREMLF90.
Run the program with YAMS (modified FSPAK). The computing time can be dramatically improved.
OPTION tol 1d-12
Tolerance (or precision) (default 1d-14) for positive definite matrix and g-inverse subroutines.
Convergence may be much faster by changing this value.
OPTION sol se
Store solutions and those standard errors.
OPTION store_pev_pec 6
Store triangular matrices of standard errors and its covariances for correlated random effects such as direct-maternal effects and random-regression effects in “pev_pec_bf90”.
y-hat and residuals will be included in “yhat_residual”.
OPTION missing -999
Specify the missing value (default 0) in integer.
OPTION constant_var 5 1 2 ...
5: effect number
1: first trait number
2: second trait number
implying the covariance between traits 1 and 2 for effect 5.
Heterogeneous residual variances for a single trait
OPTION hetres_pos 10 11
Specify the column positions of (two) covariables in the data file.
OPTION hetres_pol 4.0 0.1 0.1
Initial values of coefficients for heterogeneous residual variances using ln(a0, a1, a2, …) to make these values.
Heterogeneous residual variances for multiple traits
Convergence will be very slow with multiple trait heterogeneous residual variances
OPTION hetres_pos 10 10 11 11
OPTION hetres_pos 10 11 12 13
Specify the column positions of covariables (trait first) in the data file.
“10 10” or “10 11” could be linear for first and second traits.
“11 11” or “12 13” could be quadratic.
OPTION hetres_pol 4.0 4.0 0.1 0.1 0.01 0.01
Initial values of coefficients for heterogeneous residual variances using ln(a0, a1, a2, …) to make these values (trait first).
“4.0 4.0” are intercept for first and second traits.
“0.1 0.1” could be linear and “0.01 0.01” could be quadratic.
To transform back to the original scale, use exp(a0+a1*X1+a2*X2).
OPTION SNP_file snp
Specify the SNP file name to use genotype data.
OPTION se_covar_function <label> <function>
As an alternative of SE, calculate SD for function of (co)variances by repeated sampling of parameters estimates from their asymptotic multivariate normal distribution, following ideas presented by Meyer and Houle 2013.
A name for a particular function (e.g.,
P1 for phenotypic variance of trait 1,
H2_1 for heritability for trait 1,
rg12 for genetic correlation between traits 1 and 2, …).
A formula to calculate a function of (co)variances to estimate SD. All terms of the function should be written with no spaces.
Each term of the function corresponds to (co)variance elements and could include any random effects (G) and residual (R) (co)variances.
Notation is with reference to the effect number and the trait number (
G_eff1_eff2_trt1_trt2) that indicate the element of the (co)variance matrix for random effect
eff2 are effect numbers 1 and 2, and
trt2 are trait numbers 1 and 2.
R_trt1_trt1 indicates the element of the residual (co)variance matrix for traits 1 and 2.
Several functions could be added, with one OPTION line per function.
OPTION se_covar_function P G_2_2_1_1+G_2_3_1_1+G_3_3_1_1+G_4_4_1_1+R_1_1
OPTION se_covar_function H2d G_2_2_1_1/(G_2_2_1_1+G_2_3_1_1+G_3_3_1_1+G_4_4_1_1+R_1_1)
OPTION se_covar_function H2t (G_2_2_1_1+1.5*G_2_3_1_1+0.5*G_3_3_1_1)/(G_2_2_1_1+G_2_3_1_1+G_3_3_1_1+G_4_4_1_1+R_1_1)
OPTION se_covar_function rg12 G_2_2_1_2/(G_2_2_1_1*G_2_2_2_2)**0.5
The first function calculates the SD for the total variance for a maternal model with permanent maternal effect, where 2 and 3 are the effect number for the direct and maternal additive genetic effects respectively, and 4 is the effect number for the maternal permanent random effect.
The second function calculates the heritability for the direct component.
The third function the total heritability.
The fourth function calculates the SD of the genetic correlation between traits 1 and 2 for the direct genetic effect (effect number 2)
OPTION samples_se_covar_function <n>
Set the number of samples to calculate SE for function of (co)variances.
default value 10000
Indicate to store in file samples of (co)variances function for postprocessing (histogram, etc.)
When the covariance matrix is close to non-positive definite, the AIREMLF90 may not converge. There are two options you might want to try:
1. change the tolerance value (xx) in the option:
OPTION tol xx
to a very strict value (e.g., 1d-20) or a lenient value (1d-06).
2. use an option to use EM-REML inside AI-REML:
OPTION EM-REML xx
where xx is the number of iterations for EM-REML you expect to get a good starting value for AI-REML. After running xx rounds with EM-REML, the AIREMLF90 program will automatically switch from EM-REML to AI-REML using the last estimate from EM-REML as a starting value for AI-REML.