jarbes ChangeLog Version 2.2.1 -- June 2024 ................................................. Correction of two bugs: 1) bcmeta() function, the biased mean is mu[2] <- B =< mu.bias = mu[1]+B 2) bcmixmeta() function, the biased means of the DP for the betas[i] is mu.k[k] ~ dnorm(B, inv.var.beta) Version 2.2.0 -- November 2023 ................................................. Implementation of the bcmixmeta for the ISCB-2023 paper Implementation of the bcdpmeta for the ISCB-2023 paper Version 2.1.2 -- August 2023 ................................................... * dpmeta: Bayesian meta-analysis with Dirichlet Process with base distribution G0=N(mu.0, sigma.0^2) * dpmmeta: Bayesian meta-analysis with Dirichlet Process Mixture with base distribution G0=N(mu.0, sigma.0^2) * bcdpmeta: Bias Corrected model with Dirichlet Process. I try three different possibilities: 1) Using dnormmix in JAGS, which is extremely slow 2) Using common w_k and alpha 3) Using two different DPs, one for theta and one for beta. This is a convex combination of two DP. Version 2.1.1 -- February 2023.................................................. * jarbes does not depend of the tidyverse anymore. Only tidyr is imported. Commitments for the next version * bmeta: scale mixture random effects * diagnostic bmeta: compare Bayesian cross-validation and scale mixture weights * Vignettes: * bmeta * bcmeta * b3lmeta * metarisk * bforest: a forest plot for bmeta; bcmeta; b3lmeta. * hmr: Posterior prediction of the treatment effect to an new group of patients * Effective number of studies in bcmeta and bmeta and b3lmeta * Function: betaplot for the regression coefficients of hmr * Binomial and Poisson likelihoods in bmeta, bcmeta and b3lmeta using the arguments "family=...", "link="..." * Tweedy's formula for bias: this is to understand the different bias correction approaches (e.g. bmeta vs. bcmeta vs. b3lmeta vs. bmetareg) * Planned: Meta-regression function........................ * Function: bmetareg * Function: summary.bmetareg * Function: plot for bmetareg * Function: diagnostic for bmetareg * Function: betaplot for the regression coefficients of bmetareg * Publication bias modeling. Version 2.1.0 -- May-June-July-August 2022.................................................. * New data set: hips * Vignette for hmr() with a real example. * Important Bug correction in hmr(): From version 1.7.0 (Verde, 2019) to 1.7.4 I commented a line in the preparation of the IPD. This action generated a massive differences in the estimation of mu.phi. Now, this bug is solved. * New function: effect() for the class "hmr"". This function plots the posterior contour for effectivenes in a subgroup identify by the hmr. * The function plot.hmr() only shows the location of the baseline risk corresponding to AD and IPD. * Parameters: beta.0 and beta.1 are now alpha.0 and alpha.1. This is for consistency in the notation in the current presentation of the model. This change affect the Figures of the supplementary material of Verde (2019). * Bug correction: Odds ratio calculation with different link functions in metarisk() * Bug correction: Odds ratio calculation with different link functions in hmr() * Predictive summaries are depreciated in hmr() * Predictive summaries are depreciated in metarisk() * summary.hmr: predictive summaries omitted. * summary.metarisk: predictive summaries omitted. * The argument to run rjags is depreciated in all functions. Version 2.0.0 -- January-Febrary-March 2022 Done in version 2.0.0 ...................................... * Approximated Bayesian Cross-Validation for bmeta, b3lmeta, bcmeta. * Function: diagnostic for bcmeta (specific model and Bayesian cross-validation) * Function: diagnoistic for bmeta (Bayesian cross-validation) * Function: diagnoistic for b3lmeta (Bayesian cross-validation) * Include as argument the labels and axis information for plot functions: hmr (done) metarisk (done) bcmeta (done) bmeta (done) b3lmeta (done) * plot.bcmeta: add the arrows and the text for the distributions... * summary.b3lmeta: add the means by groups! (done) * Function: plot for bcmeta (done) * Function: plot for bmeta (done) * Function: plot for b3lmeta (done) * Function: bl3meta for three levels hierarchical meta-analysis (done) * Function: summary for bl3meta (done) * The function b3lmeta replaced the function ges * Labels (posterior-prior) in the diagnostic function of hmr (done) * Function bmeta for simple bayesian metaanalysis (done) * Function summary for bmeta (done) * Summary functions: * Include the prior parameters in the object type "hmr" (done) * Include the prior parameters in the object type "bcmeta" (done) * Include the prior parameters in the object type "metarisk" (done) * Include the prior parameters in the object type "bmeta" (done) * Include the prior parameters in the object type "b3lmeta" (done) Version 1.9.6 -- December 2021 * Function: bcmeta implements the "Bias-Corrected" Meta-analysis model * Function: summary for bcmeta * Function: plot for metarisk * Function: summary for metarisk * Function: diagnostic for metarisk * Function: plot for hmr * Function: summary for hmr * Function: diagnostic for hmr * New data frame covid19: meta-analysis of risk factors for complications and death in COVID-19 patients. * The function bcmeta replaced the function gesmix Version 1.8.0 -- Summer 2020 * The function "metamix" is not part of jarbes anymore. It will be replaced by the function "bcmeta" Version 1.7.4 -- November - December 2019 * Preparation for the paper on mixture models Version 1.7.3 -- April - May 2019 * Correction in the example of data ppvcap, metariks(..., two.by.two = TRUE,...) * Typo correction in the example of ges() "ppvipv" must be "ppvipd" Version 1.7.2 -- February-March 2019 * New function: "gesmix"" performs the finite mixture random effects bias analysis of Verde 2017 and Verde and Curcio 2019. * New arguments for the function "ges": EmBi = "Empirical Bias"standing for "penalization of observational studies" ExPe = "Explicit Penalization" for observational studies. * All if()s 've been checked. The reason was: Issue from CRAN compilation: --- failure: length > 1 in coercion to logical --- Version 1.7.1 -- December 2018 * A bug in function hmr() is fixed. Version 1.7.0 -- June 2018 * Implementation of the function "hmr" for combining aggregated data and individual participant data. * New dataset "healing": this dataset corresponds to a systematic review of aggregated data. The primary endpoint is healing without amputation in one year follow up. * New dataset "healingipd": This dataset corresponds to individual participant data for diabetic patients. Version 1.1.0 -- December 2017 * Implementation of the "ges" function for Generalized Evidence Synthesis. * Implementation of Half Cauchy priors for components of variances. * Implementation of Empirical Bias adjustement for Observational Studies. * Implementation of Penalization methods for Observationa Studies. * First prototype of the "hmr" function for combining IPD and AD data. Version 1.0.0 -- September 2017 * Implementation of the "metarisk"" function for bivariate hierarchical meta-regression of aggregated data. * Documentation improved. Version 0.5.0 -- January 2017 * Creation of data examples: * ppv.cap: PPV23 (23-valent pneumococcal polysaccharide vaccine) with 16 Randomized Clinical Trials (RCTs); outcome variable CAP (community-acquired pneumonia). * ppv.ipde: PPV23 with 3 RCTs and observational studies (5 cohorts and 3 case controls); all data types are aggregated results; outcome variable IPD (invasive pneumococcal disease). * stem.cells: 31 randomized controlled trials (RCTs) of two treatment groups of heart disease patients, where the treatment group received bone marrow stem cells and the control group a placebo treatment. * opti: one pragmatic trial, the OPTIMIZE trial, which evaluates the clinical effectiveness of a perioperative, cardiac output–guided hemodynamic therapy algorithm. And 21 small RCTs with evaluation of Risk of Bias. All are studies have aggregated data. * foot.ad: 36 RCTs which investigate adjunctive therapies vs. routine medical care in diabetic patients. The primary outcome is healing without foot amputation. There are aggregated covariates describing studies and patients characteristics. * foot.ipd: A cohort study with 260 diabetic patients, the outcome variable is healing without amputation in a followup of one year. In addition we have 14 potential risk factors.