longskat_est_model: NULL model estimation

Description Usage Arguments Value References Examples

Description

Estimating the parameters and residuals for the NULL model in LSKAT.

Usage

1
2
3
4
5
6
7
8
9
longskat_est_model( phe.long,
    phe.cov, 
    phe.time  = NULL, 
    time.cov  = 0, 
    intercept = FALSE, 
    method    = c("REML", "ML"),
    g.maxiter = 20, 
    par.init  = list(), 
    verbose   = F)

Arguments

phe.long

Phenotype matrix with m rows denoting the individuals and n columns denoting the time points, the row name indicates the individual's ID.

phe.cov

Time covariate matrix with m rows denoting individuals and x columns denoting the covariate variables, the row name indicates the individual's ID.

phe.time

Time point matrix with m rows denoting individuals and n columns denoting the time points, the row name indicates the individual's ID. If this matrix is not specified, the default matrix is generated.

time.cov

Numeric, indicating whether the time exponnents are included as extra covariates, The time points are used if 1, the time points and time squares are used if 2, and so on. The default value (0) doesn't use the time covariate.

intercept

Logical variable, indicating whether the intercept is estimated.

method

String, REML or ML are available for the parameter estimation.

g.maxiter

Numeric, the maximum count for the iterative estimation.

par.init

List, the initial values for the parameter rho, sig.a, sig.b, sig.e.

verbose

Logical variable, indicating whether some debug information can be outputted.

Value

This function returns an list object with model parameters and residuals of the NULL model which assumes there is no association between genes and longitudinal phenotypes.

The return object is a list with the following items:

par

List, model paramters as shown in below.

likelihood

Numeric, the likelihood value estimated by REML or ML.

phe.delt

Residual matrix with the row name indicating the individual's ID, the structure is same as phe.long.

phe.time

Time matrix, copied from the input parameter phe.time.

phe.cov

Covariate matrix, copied from the input parameter phe.cov.

The Model paramters: par has the following sub-items:

intercept

Logical variable copied from the input parameter, indicating the intercept is estimated.

mu

Numeric indicating the intercept value.

cov.effect

String indicating the coefficient of the covariates except intercept

sig.a

String indicating the standar deviation of individual random effects

sig.b

String indicating the standar deviation of individual-specific timede-pendent random effects.

sig.e

String indicating the standar deviation of measurement error.

rho

String indicating the corelation coefficient of covariance structure.

time.cov

Numeric, indicating whether consider times as covariate, 0 means no time effects, 1 means time effects, 2 means time effects and time square effects are included as covariates. and so on.

time.effect

Vector of numeric, the time coefficient of time effects. The 1st item is the coefficient for time effects, The 2nd item is the coefficient for time square effects and so on.

After obtaining the model parameters, please use the longskat_gene_test to test the association between gene and traits.

References

Wang Z., Xu K., Zhang X., Wu X., and Wang Z., (2016) Longitudinal SNP-set association analysis of quantitative phenotypes. Genetic Epidemiology.

Examples

1
2
3
4
5
6
7
8
## Data simulation using the default parameters
p0 <- longskat_gene_simulate();

## Estimating the model parameters and residuals
r.model0 <- longskat_est_model( p0$phe.long, p0$phe.cov, g.maxiter=3, verbose=T);

##print this model
print(r.model0);

ZWang-Lab/LSKAT documentation built on May 10, 2019, 1:55 a.m.