Skip to content

Latest commit

 

History

History
75 lines (60 loc) · 6.46 KB

NEWS.md

File metadata and controls

75 lines (60 loc) · 6.46 KB
Raw

SIBER 2.1.9

  • Problems with rendering the permille symbol on latest OSX systems has meant we have had to remove all reference to the unicode symbol "U+2023" or "\u2030" within expressions. In order to create figures with the symbol present, you will need to do so with your own code running on your own machine with code to work with your local OS and region/locale settings.
  • Re-wrote all help files to use markdown format via Roxygen2 in place of LaTeX format.
  • DOIs correctly referenced in *.Rd files using \doi{} syntax.

SIBER 2.1.8

  • Added a check that the data object passed to createSiberObject is of class data.frame only. If this is the case, it will coerce via as.data.frame(data.in) and issue a warning to this effect, recommending the coercion is done before passing to this function.
  • Updated PACKAGE help file construction per instructions from CRAN.

SIBER 2.1.7

  • removed suggests, depends or imports "tidyverse" and instead import specific packages as advised is best practice
  • added doi to DESCRIPTION per CRAN suggestion
  • added new vignette illustrating the kapow method and associated new dataset data/mongooseData.rda
  • fixed a bug in siberKapow which will hopefully pass CRAN checks re undeclared global variables. Will pick this up on next submission.

SIBER 2.1.6

  • added two vignettes illustrating how to compare populations and communities with calculation of probabilities for differences.
  • replaced package spatstat with direct importing from new sub-packages spatstat.utils and spatstat.geom

SIBER 2.1.5

  • Fixed bug in createSiberObject that meant z-scores could end up out of sync with their respective grouping variable. Thanks to Edward Doherty for pointing out this odd behaviour leading to bug discovery 2020/11/5. Ultimately the entire package needs to be recoded fully in tidyverse.
  • Fixed bug in plotSiberObject that meant both x.limits and y.limits had to be specified to invoke manual axis setting. Either or can now be specified on their own with the other reverting to the default which is the max and min values of the corresponding data axis +/- the specified ax.pad value which defaults to 1.
  • Allow users to specify a matrix of colours to define each group and each confidence boxplot uniquely using siberDensityPlot(Y, clr = ...)

SIBER 2.1.4

  • fixed "vignettes with duplicate titles" warning
  • added new vignette Plot-posterior-ellipses that shows how to plot multiple posterior ellipse draws on top of data using ggplot2.
  • fixed fitEllipse so that it now correctly inherits parms$n.chains and parms$n.thin from input
  • added ability to save the raw jags output to *.RData file to enable convergence diagnostics for each ellipse within fitEllipse. This option is set in the higher level user-specified object parms$save.output.
  • added new vignette Test-convergence to illustrate this new functionality of saving the raw jags output and testing using Gelman diagnostics.

SIBER 2.1.3

  • Contains a hotfix owing to a change to 'spatstat' detailed below
  • swapped package spatstat for spatstat.utils as per instructions from their package maintainers 23/03/2017
  • New functions siberCentroids, specificCentroidVectors and allCentoidVectors added to allow pairwise comparison of the locations of two groups using vectors. Illustrated in an accompanying vignette.

SIBER 2.1.2

  • added ability to specify custom pch point orders in plotSiberObject()
  • added ... to plotGroupHulls

SIBER 2.1.1

  • N.B. an error during uploading to CRAN meant a release was never lodged. I have moved on to v2.1.2 for the next release as above.
  • hotfix for bug in bayesianOverlap - thanks Mark Nowak for spotting this.
  • Important: install from github if you want to use bayesianOverlap for now until I can push a hotfix to CRAN. devtools::install_github("andrewljackson/SIBER", build_vingettes = TRUE)
  • examples added to maxLikOverlap() and pointsToEllipsoid().

SIBER 2.1.0

  • Added functions to calculate whether arbitrary points are inside or outside ellipses or higher dimensional ellipsoids along with illustrative vignettes.
  • Changed method of calculation of angle of ellipse with x-axis to using atan in place of asin which is a more elegant way of ensuring the sign of the returned angle is correct.
  • Small sample size correction for drawing ellipses can now be toggled using addEllipse(small.sample = TRUE, m = m) effectively meaning SEAc or SEA can be illustrated.
  • New vignette added illustrating how to calculate overlap between two ellipses. Two new functions detailed below greatly improve the ease with which this can be applied.
  • Fixed three bugs in the ellipse overlap vignette. Thanks to Sarina for pointing this out. These are no longer an issue as the new functions detailed below replace this code in the vignette, but it was helpful for me during the creation of these functions. Thanks.
  • Added a new function bayesianOverlap() that calculates the posterior overlap between ellipses fitted to two groups. Thanks to Josh Stewart for forcing my hand on this long-overdue feature.
  • Added a new function maxLikOverlap() to ease the calculation of overlap between ellipses using the ML estimated ellipses which previously required more manually coding than was ideal. Thanks Mark Nowak and SarinaJ for spotting some bugs and helping work them out.
  • added new functions ellipseInOut() and pointsToEllipsoid to enable testing of whether points lie inside or outside an n-dimensional ellipsoid, including the bivariate ellipse. These are useful for testing that the quantile prediction ellipses do indeed contain the expected number of data points from a sample. It might also be useful for assignment, identification of outliers, or measures of overlap of individual data points with other ellipses.

SIBER 2.0.3

  • Added a new vignette illustrating how to add custom ellipses to each group manually using the function addEllipse

SIBER 2.0.2

  • Bug in Group labels as character strings fixed
  • Community labels as character strings now implemented

SIBER 2.0.1

  • Group labels can now be strings and do not have to be sequential integers (#14)

SIBER 2.0

  • Major overhaul of all code and underlying fitting algorithms
  • Fitting is now via JAGS
  • Data are z-score transformed prior to fitting to improve convergence
  • Data structures have now changed from previous versions embedded within SIAR so you will have to reformat your data and write new scripts to interface with the new code