cmmb: Compositional Mediation Model for Binary Outcomes

In mbsohn/cmmb: Compositional Mediation Model for Binary Outcomes

Description

Estimate direct and indirect effects of treatment on binary outcomes transmitted through compositional mediators

Usage

cmmb(Y, M, tr, X, n.cores=NULL, n.boot=2000, ci.method="empirical",
     p.value=FALSE, ForSA=FALSE, max.rho=0.5, sig.level=0.05, FWER=FALSE,
     w=rep(1,length(Y)), prec=1e-4, max.iter=1000)

Arguments

Y	a vector of binary outcomes
M	a matrix of compositional data
tr	a vector of continuous or binary treatments
X	a matrix of covariates
n.cores	a number of CPU cores for parallel processing
n.boot	a number of bootstrap samples
ci.method	options for bootstrap confidence interval. It can be either "empirical" (default) or "percentile".
p.value	a logical value for calculating the p value. It is inactive when ci.method="percentile".
ForSA	a logical value for sensitivity analysis
max.rho	a maximum correlation allowed between mediators and an outcome
sig.level	a significance level to estimate bootstrap confidence intervals for direct and indirect effects of treatment
FWER	a logical value for family-wise error rate for direct and total indirect effects. If FWER=TRUE, the Bonferroni correct will be applied.
w	a vector of weights on samples. If measurements in a sample is more reliable than others, this argument can be used to take that information into the model.
prec	an error tolerance or a stopping criterion for the debiasd procedure
max.iter	a maximum number of iteration in the debias procedure

Note: the range of rho is not from -1 to 1 when the number of components is more than two because the correlation between them is not zero, and the range gets smaller as the number of components increases.

Value

If ForSA=FALSE,

total	contains estimated direct and total indirect effects with their confidence limits
cwprod	contains component-wise products of path coefficients with their confidence limits

If ForSA=TRUE,

total	contains estimated direct and total indirect effects with their confidence limits
cwprod	contains component-wise products of path coefficients with their confidence limits
cide.rho	contains estimated indirect effects and corresponding pointwise 95% confidence intervals, given correlations between mediators and an outcome

Author(s)

Michael B. Sohn

Maintainer: Michael B. Sohn <[e-mail]>

References

Sohn, M.B., Lu, J. and Li, H. (2021). A Compositional Mediation Model for Binary Outcome: Application to Microbiome Studies (Submitted)

Examples

## Not run: 
# Load a simulated dataset
data(cmmb_demo_data)
# Run CMM for binary outcomes
rslt <- cmmb(Y=cmmb_demo_data$Y, M=cmmb_demo_data$M,
             tr=cmmb_demo_data$tr, X=cmmb_demo_data$X)
rslt
# Plot products of component-wise path coefficients
plot_cw_ide(rslt)

## End(Not run)

mbsohn/cmmb documentation built on Dec. 21, 2021, 3:56 p.m.