sl4bayesmeta-package: Sensitivity and learning for Bayesian meta-analysis

In sl4bayesmeta: Sensitivity and learning for Bayesian meta-analysis

Description

The main function sensitivity_learning_table() provides posterior, sensitivity and learning estimates for the Bayesian normal-normal hierarchical model used for Bayesian meta-analysis under four different heterogeneity priors (half-normal, half-Cauchy, exponential, Lomax). The more advanced function sensitivity_learning_table_flexible() enables a flexible choice of several parameters and supports computation of the reference within-study standard deviation based on both the geometric mean and a weighted harmonic mean.

In order to unify notation, the heterogeneity priors are defined as scaled distributions tau ~ A_0 |X|, where A_0 is a scale parameter and X is the standard form of the distribution. The methodology implemented is proposed in Ott et al. (2020). The function pri_par_adjust_dynamic() implements the novel heterogeneity prior adjustment with respect to the relative latent model complexity (RLMC). The function pri_par_adjust_static() implements the standard approach to heterogeneity prior tail adjustment (Spiegelhalter et al. 2004).

Details

Package: sl4bayesmeta

Type: Package

Title: Sensitivity and learning for Bayesian meta-analysis

Version: 0.3-1

Date: 2020-02-18

Author: Manuela Ott [aut, cre], Malgorzata Roos [aut]

Maintainer: Manuela Ott <[e-mail]>

Depends: bayesmeta

License: GPL (>=2)

Author(s)

Manuela Ott, Malgorzata Roos Maintainer: Manuela Ott <[e-mail]>

References

Ott, M., Hunanyan, S., Held, L., Roos, M. Sensitivity quantification in Bayesian meta-analysis. Manuscript revised for Research Synthesis Methods. 2020.

Spiegelhalter, D., Abrams, K., Myles, J. (2004). Bayesian Approaches to Clinical Trials and Health-Care Evaluation. John Wiley & Sons, Ltd.

Examples

# Acute Graft rejection (AGR) data analyzed in Friede et al. (2017), 
# Sect. 3.2, URL: https://doi.org/10.1002/bimj.201500236
df <- data.frame(y = c(-2.310, -1.258), # log-odds-ratio
                  sigma = c(0.599, 0.642), # SE(log-odds-ratio)
                  labels = c(1:2))
  
# compute posterior, sensitivity and learning estimates for AGR data
# warning: it takes ca. 5-10 minutes to run this function 
# on the above data set!
sensitivity_learning_table(df)

# dynamic prior adjustement based on RLMC
pri_par_adjust_dynamic(df=df, rlmc=0.25)
# static 5 % prior tail adjustement with reference threshold UU=1
pri_par_adjust_static(UU=1)

sl4bayesmeta documentation built on Feb. 18, 2020, 3:02 p.m.