Added the option to specify a dependence structure in
adjPrevSensSpecCI (i.e. argument Sigma allowed
and passed through to bootComb).
Added the option to specify a seed value for
adjPrevSensSpecCI.
Added the option for bootstrapped values to be returned for
adjPrevSensSpecCI.
Added a note to the documentation files for getPoisFromCI and
getNegBinFromCI - these two functions are unlikely to be of much use
within bootComb (where parameters are necessarily assumed to be
continuous).
Minor fix to one example.
bootComb 1.1.1
Reduced the number of bootstrap samples for the example used for the
bootComb example as CRAN gave an error that examples take
over 10s to run.
Added a seed value for the examples for the bootComb
function.
bootComb 1.1.0
Support added for dependent / correlated parameters (using Gaussian
copulas to specify the dependence). This is done by specifying a
correlation matrix (new input argument Sigma).
Added an alternative, and simpler, way of calling the main package
function bootComb (new input arguments
distributions, qLowVect, qUppVect
and alphaVect).
returnBootVals returns now also the values for the
individual parameters that get combined. This change allows checking the
dependence structure that was specified.
A random seed can now be specified as an input argument to the
bootComb function (argument seed).
Update to README.md (example of alternative specification and
citation info to avoid hard-coded reference to version number).
bootComb 1.0.2
Citation information updated now the IJE paper is out. Please cite
https://doi.org/10.1093/ije/dyab049 when using the package.
Updated the function adjPrevSensSpecCI for better
behaviour in the case where the provided point estimates for the
prevalence, sensitivity and specificity result in an impossible value
(adjusted prevalence point estimate <0 or >1). The adjusted point
estimates and boundaries of CI have now been updated to avoid impossible
values and be consistent with each other (e.g. if the adjusted point
estimate would be -0.000123, the lower boundary from the bootstrapping
0.000451 and the upper boundary 0.112345, then the reported point
estimate and CI will be 0 [0,0.112345]).
bootComb 1.0.1
Fixed a bug with the internal functions
identifyBetaPars, identifyNormPars,
identifyPoisPars, identifyNegBinPars,
identifyGammaPars, identifyExpPars whereby the
alpha parameter did not get passed through.
bootComb 1.0.0
With the inclusion of additional distributions functions that are
supported, this is considered to be the first mature version of the
package. Some changes for consistency of coding and better function
naming are also minimally not backward compatible, hence the upgrade to
version number 1.0.0.
Added cranlogs badges to README.md
Changed the way parameters are found for the Poisson distribution.
Rather than quantile matching, the CDF is matched (as this will be a
continuous function rather than a step function in the way the quantile
function is).
Added support for negative binomial, gamma and exonential
distributions in addition to the beta, normal and Poisson distributions.
Finding the best-fit negative binomial distribution proceeds similar to
the procedure for the Poisson distribution as this is another discrete
distribution and hence the quantile function is a step function (albeit
a two-dimensional step function in this case).
Renamed the (internal only) function ‘ssPoisPar’ to ‘ssPoisPars’ for
consistency with other distributions.
Renamed arguments ‘pLow’ and ‘pUpp’ from getBetaFromCI() to ‘qLow’
and ‘qUpp’ for consistency. Updated code from other functions where
these parameters were specified.
Default method for interval computation changed from
hdi to quantile as the latter will always be
correct and the former may be wrong if the bootstrap sample of combined
parameter values is severely multimodal.
Dependency on package HDInterval turned into a
“Suggests” rather than an “Imports” dependency with
bootComb falling back on method='quantile' if
HDInterval is not available.
Small update to the example from the README.md file.
Clarified the names for the functions to run the 2 example
simulations.
bootComb 0.2.0
First version accepted & published on CRAN.
Various bug fixes.
Second simulation scenario added (for adjusted prevalence estimation
situation).
Ability to specify number of optimisation iterations for
identifyBetaPars
identifyBetaPars is no longer an exported function;
strictly internal
Added getBetaFromCI, a function that returns beta
distribution functions from a given confidence interval (and in
particular a random sampling function)
Added support for normal and Poisson distributions in addition to
the beta distribution
Changed the organisation of code within the source R code files
Implemented changes requested from the submission of v0.1.0 to CRAN
(references in DESCRIPTION file, TRUE/FALSE instead of T/F, restoring
user parameters after doing plots)
bootComb 0.1.0
First version to be submitted to CRAN (but rejected).