README.md

In vandy10s/mnma: Multivariate Network Meta-Analysis (NMA) within a Bayesian Framework

Hyunsoo Hwang & Stacia DeSantis

mnma

The goal of mnma package is to make it easy to run Bayesian multivariate network meta-analysis (NMA) method by Efthimiou et al. (Biostatistics, 2015) using OpenBUGS. Although there are many packages for the univariate NMA, there is no package available for multivariate NMA due to its complexity. To run bugs command in this package, OpenBUGS has to be pre-installed in the local computer.

Installation

You can install mnma from github with:

# install.packages("devtools")
devtools::install_github("vandy10s/mnma")

Example

This is a basic example which shows you how to use functions in this pacakge

1. Split dataset into outcome1, outcome2, and treatments.

First 2 (or 3) columns in the outcome contain the information about number of events for each arm and the rest 2 (or 3) columns contain the sample size for each arm. Treatments column contains maximum 3 treatment arms for each study.

library(mnma)
data(acute)
outcome1 <- acute[,c(1:6)]     # outcome1
outcome2 <- acute[,c(7:12)]    # outcome2
TX    <- acute[,13:15]

2. Convert data

mnma.model function converts outcome data into log odds ratio (OR) inputs and arrange them for multivariate NMA.

data <- mnma.model(outcome1,outcome2,TX, ref=1, parameters="simple")

3. Run multivariate NMA using mnma.run function and check the result

Adjust options in the function in the similar way as OpenBUGS. Type res$summary after running the function to check the summarized results. The BUGS code is executed automatically depending on the condition of inputs.

res <- mnma.run(data, n.iter=3000, n.burnin=1000)

4. Similarly, run univariate NMA using unma.run function and compare results with those from multivariate NMA model.

vandy10s/mnma documentation built on May 19, 2019, 8:24 a.m.