ipdlme: IPD Linear Mixed-Effects Models from Aggregate Data

In ipdmeta: Tools for subgroup analyses with multiple trial data using aggregate statistics

Description

Fits the IPD linear mixed-effects model for a treatment and single categorical covariate and their interaction using study-level summary statistics.

Usage

ipdlme(n, y, s2, max.iter = 100, tol = 1e-10,equal.residual.var=TRUE) 

Arguments

n	a list ordered by trial with each element a data frame having the columns trt and ctrl and the row the sample sizes in each group for the covariate of interest, e.g. males/females, never/former/current smokers.
y	a list ordered by trial with each element a data frame having the columns trt and ctrl and the row the average outcome (the response variable of the model) in each group for the covariate of interest.
s2	A vector of the study overall sample variances for the outcome or a list of the treatment-subgroup sample variances following the structure of n and y.
max.iter	An integer indicating the maximum number of iterations to perform in the optimization algorithm
tol	A numeric value that is used for the termination rule (see details).
equal.residual.var	logical, indicates whether the residual variances for the outcomes are assumed constant between studies or are heterogeneous

Details

The last group in the covariates, designated by the final row in the list arguments, will be treated as the reference variable.

The order of the list components should agree for all arguments.

The optimization algorithm performs maximum likelihood estimation for the IPD LME model. Given the treatment indicator z the factor x and the study variable study, the fitted model is equivalent to y~x*z+(z|study) fitted with lmer with the option REML=FALSE.

The termination rule looks at the percentage change for the fixed and study random effects after the first iteration. Convergence is declared at the first iteration for which the maximum percentage change is less than tol.

Value

An object of class ipdlme-class, for which a number of methods are available.

Author(s)

Stephanie A. Kovalchik <[e-mail]>

See Also

The ipdlme-class

Examples

data(regress_age)

# Homogeneous residual variance
metafit <- ipdlme(n,Y,S2)

summary(metafit)

fixef(metafit)
confint(metafit)

ranef(metafit)

labs <- paste("Center",c(1,10,11,2:9))
plot(metafit, y=labs)

#Heterogeneous residual variance
metafit <- ipdlme(n,Y,S2,equal=FALSE)

summary(metafit)

fixef(metafit)
confint(metafit)

ranef(metafit)

labs <- paste("Center",c(1,10,11,2:9))
plot(metafit, y=labs)

ipdmeta documentation built on May 2, 2019, 3:29 p.m.