free.estimate: Estimate parameters with free approach

In ZIBBSeqDiscovery: Zero-Inflated Beta-Binomial Modeling of Microbiome Count Data

Description

Estimate unknown parameters with free approach.

Usage

free.estimate(m, p, q, n, betastart, bvarstart, psi.start, eta.start, 
              Y, X, Y.c, ziMatrix)

Arguments

m	Number of OTUs.
p	Number of covariates for count model (e.g., beta-binomial).
q	Number of covariates for zero model.
n	Number of samples.
betastart	Matrix of estimated betas, which are the effects/coefficients for the count model, with dimension p by m. It is used as initial values for the optimization procedure to estimate betas.
bvarstart	Matrix of variance of estimated betas with dimension p by m.
psi.start	Estimated vector of logit of overdispersion parameters with length m. And psi.start will be used as initial values for the optimization procedure to estimate psi.
eta.start	Matrix of estimated etas, which are the effects/coefficients for the zero model, with dimension q by m. It is used as initial values for the optimization procedure to estimate etas.
Y	Count matrix with dimension n by m.
X	The design matrix (n by p, p is the number of covariates) for the count model (e.g., beta-binomial), and intercept is included.
Y.c	Vector of library size with length n.
ziMatrix	The design matrix (n by q) for the zero model, and intercept is included.

Value

betahat	Estimation matrix of beta (p by m).
bvar	Estimation matrix of the variance of estimated betahat (p by m).
psi	Estimation vector of the logit of the overdispersion parameters (with length m).
eta	Estimation matrix of eta (q by m).

Author(s)

Tao Hu, Yihui Zhou

ZIBBSeqDiscovery documentation built on May 2, 2019, 4:21 a.m.