revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.5.3

revdbayes 1.5.3

Bug fixes and minor improvements

• The issue described at RcppCore/Rcpp#1287 has been fixed to avoid WARNINGs from CRAN checks on some platforms. Thank you to Dirk Eddelbuettel for providing the fix so quickly!

• Fixed issues with the incorrect use of \itemize in some Rd files.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.5.2

revdbayes 1.5.2

Bug fixes and minor improvements

• The unnecessary C++11 specification has been dropped to avoid a CRAN Package Check NOTE.

• README.md: Used app.codecov.io as base for codecov link.

• Create the help file for the package correctly, with alias revdbayes-package.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.5.1

revdbayes 1.5.1

Bug fixes and minor improvements

• Removed evdbayes:: completely from tests/testthat/test-inf_priors.R to avoid a WARNING in the r-oldrel-windows-ix86+x86_64 CRAN check in checking for unstated dependencies in 'tests' ...

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.5.0

revdbayes 1.5.0

New features

• When calling predict.evpost(object, ...), if object$model = "bingp" and object$sim_vals has a third column named "theta" containing a posterior sample for the extremal index, then predictive inferences incorporate this posterior sample. This feature is introduced to facilitate the predict.blite() function in the upcoming version 1.1.0 of the lite package.

Bug fixes and minor improvements

• Dependence on the previously suggested package evdbayes has been removed because evdbayes has been archived on CRAN.

• WARNINGs in the CRAN package check results, like "init.c:120:52: warning: a function declaration without a prototype is deprecated in all versions of C [-Wstrict-prototypes] extern SEXP _revdbayes_RcppExport_registerCCallable();" have been avoided.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.4.9

revdbayes 1.4.9

New features

• The function kgaps_post() can now accept a data argument that

  • is a matrix of independent subsets of data, such as monthly or seasonal time series from different years,
  • contains missing values, that is, NAs.

• A new function dgaps_post() produces random samples from a posterior distribution for the extremal index based on what we call the D-gaps model of Holesovsky, J. and Fusek, M. Estimation of the extremal index using censored distributions. Extremes 23, 197–213 (2020). doi: 10.1007/s10687-020-00374-3. dgaps_post() has the same functionality as kgaps_post().

Bug fixes and minor improvements

• The print method print.evpost avoids printing a long list by printing only the original function call.

• The default value of inc_cens in kgaps_post() is now inc_cens = TRUE.

• In the (extremely rare) cases where grimshaw_gp_mle() errors or returns an estimate for which the observation information is singular, a fallback function is used, which maximises the log-likelihood using stats::optim()

• In the generalised Pareto example in the introductory vignette, it is now noted that for the Gulf of Mexico data a threshold set at the 95% threshold results in only a small number (16) of threshold excesses.

• In the GP section of the introductory vignette a link is given to the binomial-GP analysis in the Posterior Predictive Extreme Value Inference vignette.

• In the introductory vignette: corrected references to plots as "on the left" when in fact they were below, and corrected "random example" to "random sample".

• The microbenchmark results have been reinstated in the "Faster simulation using revdbayes" vignette.

• Activated 3rd edition of the testthat package

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.3.9

revdbayes 1.3.9

Bug fixes and minor improvements

• Tests in test-gp.R, test-gev.R and test-bingp.R have been modified to avoid errors in the upcoming new release of the testthat package.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.3.8

revdbayes 1.3.8

Bug fixes and minor improvements

• The functions grimshaw_gp_mle(), gp_pwm() and gp_lrs() are now exported, so that the rust package can access them using :: not :::.

• The hyperlinks to the Grimshaw (1993) paper in the documentation to grimshaw_gp_mle() and set_prior() have been corrected.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.3.7

revdbayes 1.3.7

Bug fixes and minor improvements

• Fixed a bug in dgp() that produced an incorrect value for the log-density (log = TRUE) when shape is negative and very close to zero and x = -1/shape.

revdbayes 1.3.6

Bug fixes and minor improvements

• Use inherits() to check the class of objects returned from try(), rather than class().

• pkgdown documentation at https://paulnorthrop.github.io/revdbayes/

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.3.5

revdbayes 1.3.5

Bug fixes and minor improvements

• The d/p/q function for the GEV and GP distributions now handle correctly cases where the input has length 0 and/or is NA and inputs Inf and -Inf.

revdbayes: Ratio-of-Uniforms Sampling for Bayesian Extreme Value Analysis version 1.3.4

revdbayes 1.3.4

New features

• In set_bin_prior() the user can specify their own prior for the binomial probability, by providing an R function.

Bug fixes and minor improvements

• In rpost() and rpost_rcpp() an error is thrown if the prior and the model are not compatible. Previously a warning was given.

• The penultimate example in the documentation for set_prior() has been corrected by adding model = "gp". The default model = "gev"` is not appropriate here because the prior is set up for the GP model.

• (This is an amendment to the third minor improvement in the NEWS for v1.3.3.) In rpost() and rpost_rcpp() an error is thrown if the input threshold thresh is lower than the smallest observation in data. This is only checked when model = "bingp" or model = "pp". This not checked when model = "gp" because the user may legitimately supply only threshold excesses. (Many thanks to Leo Belzile for spotting this.)