MAI2: Multivariate Average Information Algorithm

Description

This function is used internally in the function mmer when MORE than 1 variance component needs to be estimated through the use of the average information (MAI) algorithm and the MVM argument is set to TRUE.

Usage

MAI2(Y,X=NULL,ZETA=NULL,init=NULL,maxcyc=20,tol=1e-3, tolparinv=1e-6, draw=TRUE, silent=FALSE, constraint=FALSE,EIGEND=FALSE,forced=NULL)

Arguments

a matrix or data frame of response variables

an incidence matrix for fixed effects.

ZETA

incidence matrices and var-cov matrices for random effects. This works for ONE OR MORE random effects. This needs to be provided as a 2-level list structure. For example: ' ETA <- list( A=list(Z=Z1, K=K1), B=list(Z=Z2, K=K2), C=list(Z=Z3, K=K3) ) ' makes a 2 level list for 3 the random effects A, B and C, stored in a variable we call ETA. The general idea is that each random effect is a list, i.e. A=list(Z=Z1, K=K1) where Z is the incidence matrix and K the var-cov matrix for the random effect, if K is not provided is assumed an identity matrix conferring independence. ' PLEASE remember to use the names Z and K for all random effects when you provide your matrices, that's the only way the program distinguishes between a Z or a K matrix. ' To provide extra detail, I'll rephrase it; when moving to situations of more than one random effect, you need to build a list for each random effect, and at the end everything gets joined in a list as well (BGLR type of format). Is called a 2-level list, i.e. A=list(Z=Z1, K=K1) and B=list(Z=Z2, K=K2) refers to 2 random effects and they should be put together in a list: ' ETA <- list( A=list(Z=Z1, K=K1), B=list(Z=Z1, K=K1) ) ' Now you can fit your model as: ' mod1 <- mmer(y=y, Z=ETA) ' You can see the examples at the bottom to have a clearer idea how to fit your models.

init

vector of initial values for the variance components. By default this is NULL and variance components are estimated by the method selected, but in case the user want to provide initial values this argument is functional.

maxcyc

Maximum number of cycles allowed. Default value is 50. A warning is output to the screen if this is reached before convergence.

draw

a TRUE/FALSE value indicating if a plot of updated values for the variance components and the likelihood should be drawn or not. The default is TRUE. COMPUTATION TIME IS SMALLER IF YOU DON'T PLOT SETTING draw=FALSE

tol

Convergence criteria. If the change in residual log likelihood for one cycle is less than 10 x tol the algorithm finishes. If each component of the change proposed by the Newton-Raphson is lower in magnitude than tol the algorithm finishes. Default value is 1e-4.

tolparinv

tolerance parameter for matrix inverse

silent

a TRUE/FALSE value indicating if the function should draw the progress bar while working or should not be displayed. The default is FALSE, which means is not silent and will display the progress bar.

constraint

a TRUE/FALSE value indicating if the function should apply the boundary constraint indicating that variance components that are zero should be removed from the analysis and variance components recalculated.

EIGEND

a TRUE/FALSE value indicating if an eigen decomposition for the additive relationship matrix should be performed or not. This is based on Lee (2015). The limitations of this method are: 1) can only be applied to one relationship matrix 2) The system needs to be squared and no missing data is allowed (then missing data is imputed with the median). The default is FALSE to avoid the user get into trouble but experimented users can take advantage from this feature to fit big models, i.e. 5000 individuals in 555 seconds = 9 minutes in a MacBook 4GB RAM.

forced

a list of values for variance-covariance components to be used if the user wants to force those values.

Value

If all parameters are correctly indicated the program will return a list with the following information:

Details

This algorithm is based on Tunnicliffe (1989), it is based on REML. This handles models of the form:

y = Xb + Zu + e

b ~ N[b.hat, 0] ............zero variance because is a fixed term

u ~ N[0, K*sigma(u)] .......where: K*sigma(u) = G

e ~ N[0, I*sigma(e)] .......where: I*sigma(e) = R

y ~ N[Xb, var(Zu+e)] ......where;

var(y) = var(Zu+e) = ZGZ+R = V which is the phenotypic variance

The function allows the user to specify the incidence matrices with their respective variance-covariance matrix in a 2 level list structure. For example imagine a mixed model with the following design:

fixed = only intercept.....................b ~ N[b.hat, 0]

random = GCA1 + GCA2 + SCA.................u ~ N[0, G]

where G is:

|K*sigma(gca1).....................0..........................0.........| |.............0.............S*sigma(gca2).....................0.........| = G

|.............0....................0......................W*sigma(sca)..|

The likelihood function optimized in this algorithm is:

logL = -0.5 * (log( | V | ) + log( | X'VX | ) + y'Py

where: | | refers to the derminant of a matrix

References

Tunnicliffe W. 1989. On the use of marginal likelihood in time series model estimation. JRSS 51(1):15-27.

Covarrubias-Pazaran G (2016) Genome assisted prediction of quantitative traits using the R package sommer. PLoS ONE 11(6): doi:10.1371/journal.pone.0156744

Examples

Run this code

####=========================================####
#### For CRAN time limitations most lines in the 
#### examples are silenced with one '#' mark, 
#### remove them and run the examples
####=========================================####
data(CPdata)
CPpheno <- CPdata$pheno
CPgeno <- CPdata$geno
### look at the data
head(CPpheno)
CPgeno[1:5,1:5]
## fit a model including additive and dominance effects
Y <- CPpheno
Za <- diag(dim(Y)[1])
A <- A.mat(CPgeno) # additive relationship matrix
####================####
#### ADDITIVE MODEL ####
####================####
ETA.A <- list(add=list(Z=Za,K=A))
#ans.A <- MAI2(Y=Y, ZETA=ETA.A)
#ans.A$var.comp

Run the code above in your browser using DataLab