jarbes (version 2.2.5)

bchmr: Bias-Corrected Bayesian Nonparametric Model to combine aggregated and individual participant data for cross design synthesis.

Description

This function performers a Bayesian cross design synthesis. The function fits a hierarchical meta-regression model based on a BC-BNP model

Usage

bchmr(
  data,
  two.by.two = TRUE,
  dataIPD,
  re = "normal",
  mean.mu.1 = 0,
  mean.mu.2 = 0,
  mean.mu.phi = 0,
  sd.mu.1 = 1,
  sd.mu.2 = 1,
  sd.mu.phi = 1,
  sigma.1.upper = 5,
  sigma.2.upper = 5,
  sigma.beta.upper = 5,
  mean.Fisher.rho = 0,
  sd.Fisher.rho = 1/sqrt(2),
  nr.chains = 2,
  nr.iterations = 10000,
  nr.adapt = 1000,
  nr.burnin = 1000,
  nr.thin = 1
)

Value

This function returns an object of the class "bchmr". This object contains the MCMC output of each parameter and hyper-parameter in the model, the data frame used for fitting the model, and further model outputs.

The results of the object of the class hmr can be extracted with R2jags. In addition a summary, a print and a plot function are implemented for this type of object.

Arguments

data

Aggregated data results: a data frame where the first four columns containing the number of events in the control group (yc), the number of patients in the control group (nc), the number of events in the treatment group (yt) and the number of patients in the treatment group (nt). If two.by.two = TRUE a data frame where each line contains the trial results with column names: yc, nc, yt, nt.

two.by.two

If TRUE indicates that the trial results are with names: yc, nc, yt, nt.

dataIPD

Individual participant data: a data frame where the first column is the outcome variable and the other columns represent individual participant charachteristics.

re

Random effects distribution for the resulting model. Possible values are normal for bivariate random effects and sm for scale mixtures.

mean.mu.1

Prior mean of baseline risk, default value is 0.

mean.mu.2

Prior mean of treatment effect, default value is 0.

mean.mu.phi

Prior mean of the bias parameter which measures the difference between the baseline mean mu.1 and the intercept parameter of the logistic regression of the individual participant data. The defalut vaule is 0.

sd.mu.1

Prior standard deviation of mu.1, default value is 1. The default prior of mu.1 is a logistic distribution with mean 0 and dispersion 1. The implicit prior for mu.1 in the probability scale is a uniform between 0 and 1.

sd.mu.2

Prior standard deviation of mu.2, default value is 1. The default prior of mu.2 is a logistic distribution with mean 0 and dispersion 1. The implicit prior for mu.2 in the probability scale is a uniform between 0 and 1.

sd.mu.phi

Prior standard deviation of mu.phi, default value is 1.

sigma.1.upper

Upper bound of the uniform prior of sigma.1, default value is 5.

sigma.2.upper

Upper bound of the uniform prior of sigma.2, default value is 5.

sigma.beta.upper

Upper bound of the uniform prior of sigma.beta, default value is 5.

mean.Fisher.rho

Mean of rho in the Fisher scale, default value is 0.

sd.Fisher.rho

Standard deviation of rho in the Fisher scale, default value is 1/sqrt(2).

nr.chains

Number of chains for the MCMC computations, default 5.

nr.iterations

Number of iterations after adapting the MCMC, default is 10000. Some models may need more iterations.

nr.adapt

Number of iterations in the adaptation process, default is 1000. Some models may need more iterations during adptation.

nr.burnin

Number of iteration discarded for burnin period, default is 1000. Some models may need a longer burnin period.

nr.thin

Thinning rate, it must be a positive integer, the default value 1.

Details

The model is experimental and under construction for the version 2.2.5 (March 2025)

References

Verde, P. E. (2019) Learning from Clinical Evidence: The Hierarchical Meta-Regression Approach. Biometrical Journal. Biometrical Journal; 1-23.

Verde, P.E., and Rosner, G.L. (2025), A Bias-Corrected Bayesian Nonparametric Model for Combining Studies With Varying Quality in Meta-Analysis. Biometrical Journal., 67: e70034. https://doi.org/10.1002/bimj.70034

Examples

if (FALSE) {
library(jarbes)

data("healing")
AD <- healing[, c("y_c", "n_c", "y_t", "n_t")]

data("healingipd")

IPD <- healingipd[, c("healing.without.amp", "PAD", "neuropathy",
"first.ever.lesion", "no.continuous.care", "male", "diab.typ2",
"insulin", "HOCHD", "HOS", "CRF", "dialysis", "DNOAP", "smoking.ever",
"diabdur", "wagner.class")]

mx1 <- bchmr(AD, two.by.two = FALSE,
           dataIPD = IPD,
           re = "normal",
           sd.mu.1 = 2,
           sd.mu.2 = 2,
           sd.mu.phi = 2,
           sigma.1.upper = 5,
           sigma.2.upper = 5,
           sigma.beta.upper = 5,
           sd.Fisher.rho = 1.25,
           df.estimate = FALSE,
           df.lower = 3,
           df.upper = 10,
           nr.chains = 1,
           nr.iterations = 1500,
           nr.adapt = 100,
           nr.thin = 1)

print(mx1)


# End of the examples.

}