Internal functions

add_genotypes(mme::MME,file,G;separator=' ',header=true,center=true,G_is_marker_variance=false,df=4.0)

• Get marker informtion from a genotype file (same order as the phenotype file).

• G defaults to the genetic variance with degree of freedom df=4.0.

• File format:

Animal,marker1,marker2,marker3,marker4,marker5
S1,1,0,1,1,1
D1,2,0,2,2,1
O1,1,2,0,1,0
O3,0,0,2,1,1

same to add_genotypes

build_model(model_equations::AbstractString,R;df::Float64=4.0)

• Build models from model equations with residual varainces R and degree of freedom for residual variance df defaulting to 4.0.

• By default, all variabels in model_equations are fixed and factors. Set variables to be covariates or random using functions set_covariate() or set_random().

#single-trait
model_equations = "BW = intercept + age + sex"
R               = 6.72
models          = build_model(model_equations,R);

#multi-trait
model_equations = "BW = intercept + age + sex
                   CW = intercept + litter";
R               = [6.72   24.84
                   24.84  708.41]
models          = build_model(model_equations,R);

get_pedigree(pedfile::AbstractString;header=false,separator=' ')

• Get pedigree informtion from a pedigree file.

• File format:

a 0 0
b 0 0
c a b
d a c

outputMCMCsamples(mme::MME,trmStr::AbstractString...)

Get samples for specific variables.

runMCMC(mme,df;Pi=0.0,estimatePi=false,chain_length=1000,starting_value=false,printout_frequency=100,missing_phenotypes=false,constraint=false,methods="conventional (no markers)",output_samples_frequency::Int64 = 0)

Run MCMC (marker information included or not) with sampling of variance components.

• available methods include "conventional (no markers)", "BayesC0", "BayesC", "BayesCC","BayesB".

• missing_phenotypes

• Pi for single-trait analyses is a number; Pi for multi-trait analyses is a dictionary such as Pi=Dict([1.0; 1.0]=>0.7,[1.0; 0.0]=>0.1,[0.0; 1.0]=>0.1,[0.0; 0.0]=>0.1),

  • if Pi (Π) is not provided in multi-trait analysis, it will be generated assuming all markers have effects on all traits.

• save MCMC samples every output_samples_frequency iterations

• starting_value can be provided as a vector for all location parameteres except marker effects.

• print out the monte carlo mean in REPL with printout_frequency

• constraint=true if constrain residual covariances between traits to be zero.

set_covariate(mme::MME,variables::AbstractString...)

• set variables as covariates; mme is the output of function build_model().

#After running build_model, variabels age and year can be set to be covariates as
set_covariate(models,"age","year")
#or
set_covariate(models,"age year")

set_random(mme::MME,randomStr::AbstractString,G;df=4)

• set variables as i.i.d random effects with variances G whose degree of freedom df defaults to 4.0.

#single-trait (example 1)
model_equation  = "y = intercept + litter + sex"
model           = build_model(model_equation,R)
G               = 0.6
set_random(model,"litter",G)

#multi-trait
model_equations = "BW = intercept + litter + sex
                   CW = intercept + litter + sex"
model           = build_model(model_equations,R);
G               = [3.72  1.84
                   1.84  3.41]
set_random(model,"litter",G)

set_random(mme::MME,randomStr::AbstractString,ped::Pedigree, G;df=4)

• set variables as random polygenic effects with pedigree information ped, variances G whose degree of freedom df defaults to 4.0.

#single-trait (example 1)
model_equation  = "y = intercept + Age + Animal"
model           = build_model(model_equation,R)
ped             = get_pedigree(pedfile)
G               = 1.6
set_random(model,"Animal Animal*Age", ped,G)

#single-trait (example 2)
model_equation  = "y = intercept + Age + Animal + Animal*Age"
model           = build_model(model_equation,R)
ped             = get_pedigree(pedfile)
G               = [1.6   0.2
                   0.2  1.0]
set_random(model,"Animal Animal*Age", ped,G)

#multi-trait
model_equations = "BW = intercept + age + sex + Animal
                   CW = intercept + age + sex + Animal"
model           = build_model(model_equations,R);
ped             = get_pedigree(pedfile);
G               = [6.72   2.84
                   2.84  8.41]
set_random(model,"Animal", ped,G)

showMME(mme::MME,df::DataFrame)

• Show left-hand side and right-hand side of mixed model equations (no markers).

solve(mme::MME,df::DataFrame;solver="default",printout_frequency=100,tolerance = 0.000001,maxiter = 5000)

• Solve the mixed model equations (no marker information) without estimating variance components.

Available solvers includes default,Jacobi,GaussSeidel,Gibbs sampler.

Construct mixed model equations with

incidence matrix: X ; response : ySparse; left-hand side : mmeLhs ; right-hand side : mmeLhs ;