bamdit ChangeLog
Commitments for the next version
* New function "bforest" for forest plot of posteriors of
sensitivity and specificity.
* Modeling function to analyze comparative test
* Meta-regression for sensitivity and specificity
* Implement diagnostic function with approximate Bayesian cross-validation
* Implement diagnostic functions and depreciate the weights plots
Version 3.3.4 -- March 2022
* New function "hyper.posterior" matrix-plot for posteriors of hyper parameters.
Version 3.3.3 -- August 2021
* New "summary" function for metadiag().
* The default option for re.model = "SeSp".
* New argument "jag.seed" to make results replicable in metadiag().
* New warning message from "metadiag()" when the number of studies is less than 6.
Version 3.3.2 -- November 2020
* Function "metadiag()": the argument "r2jags" has been depreciated.
* Several bugs in ploting functions had been fixed.
* New data frame "skin" for deep-learning diagnostic tests.
Version 3.3.1 -- July 2019
* Some keywords are replaced.
* Groupping variable "group" is the name (as character) of the data frame
column. This change applies to the following function: plotdata; plotw;
plotcompare.
* Bug in the x-y limits in plotdata() function corrected.
* Bug in the x-y limits in the plot.metadiag() function corrected.
* The function plotcompare() understands the two.by.two argument from metadiag().
* Title argument for the function plotw().
* The metadiag() function returns the posteriors of "se"" and "sp" for each study.
* The metadiag() function collects the studies' names.
* The dataframe "ct" has the author and year information.
* New data frame "diabetes" for comparative studies.
* New data frame "rapt" for comparative studies.
Version 3.2.1 -- September 2018.
* The CITATION file corresponds to the JSS paper.
Version 3.2.0 -- August 2018.
* Corrections in the documentation.
* Link the package to the JSS paper.
* Function plotdata(), the argument max.size is active.
Version 3.1.0 -- May 2017
* In metadiag() and plotdata() we allow that the data format is given
as 2x2 table with columns' names: TP, FP, TN, FN.
* A bug is fixed in the calcuation of the posterior distribution of BAUC.
* The BSROC is calculated for parametrization: re.model = "SeSp".
* New Summary function for metadiag.
* New Print function for metadiag.
Version 3.0.0 -- August 2016
* Implementation of S3 OOP in bamdit.
* metadiag is now a generic function.
* The argument re.model in metadiag allows to specify random effects on sensitivities
and specificities.
* Priors: hyper parameters mu.S and mu.D based on logistic distributions with
mean = 0 and scale = 1
* Priors for the degrees of freedom parameter: df truncated exponential.
* The function metadiag calculates the posterior probabilities of outliers.
* New functions: print; summary and plot for metadiag objects.
* The new plot function summarizes data and model predictions.
* The BSROC is only displayed in the range of the observed fpr. If this range is less
than 20% the function gives a warning.
* We added further documentation and new examples.
* Further documentation.
* More validation in input arguments.
Version 2.0.1 -- June 2015
* minor typos fixed
Version 2.0 -- June 2015
* The function bsroc() implements te Bayesian SROC curve
* Bayesian Predictive surface added (BPS)
* Calculation of the Bayesian area under the curve (BAUC)
* Migration of all graphical functions, they use ggplot
* The package "coda" is not required
Version 1.9 -- 2014
* The function metadiag() has a new implementation with blueprint() function within metadiag()
* option of using rjags or R2jags in metadiag()
* conflict of evidence analysis in metadiag() by splitting the variable w in w1 and w2
* Added a function for simulation of data (sim.meta)
Version 1.4 -- 2014-07-08
* Improvement in the documentation
Version 1.3 -- 2013-03-15
* New version of the function metadiag(). This version has main changes:
* 1) The number of degrees of freedom in the model are fixed to a default value
* 2) The Wishart prior distribution of the variance covariance matrix is replaced
* by a conditional model where the priors are given to individual components.
* 3) The priors of the variance covariance distribution are design to avoid boundary
* problems in the parameter space.
* The three adaptation trials are omited in metadiag()
Version 1.2 -- 2012-07-31
* Added more graphical functions
* We added meta-analysis examples data.
* Improved internal model functions
Version 1.1.1 -- 2011-12-08
* Examples do not run during testing.
Version 1.1 -- 2011-08-30
* Weights are return from the bamdit function
when random effects are scale mixed.
* Change warning messages when model fails to adapt in JAGS.
Version 1.0.1 -- 2011-08-09
* Nothing has to be written to disk anymore.
* Models are compiled / adapted now with "first of three"
due to an issue where sometimes models don't adapt.
* Added ChangeLog
* Corrected model adaption process from three to two stages
Version 1.0 -- 2011-08-07
* Initial release