Changelog

Source: NEWS.md

rbms 1.3.1

New features

• flight_curve() gains parallel and n_cores to fit the flight curves of the years in parallel. The random streams of the workers are seeded from the session, so the result is reproducible with set.seed().

Changes

• flight_curve_optimised() and get_nm_optimised() are deprecated: they gave the same results as flight_curve() and get_nm(), which they now call with a warning.
• flight_curve_multi(), fit_bam_multi() and plot_flight_curve_multi() (one bam() model for all years) are removed from the package for further development; they return an error, the previous code is available in rbms 1.3.0.
• flight_curve() checks all the required columns of ts_season_count (only the first one was checked).
• flight_curve() with KeepModel = FALSE and KeepModelData = TRUE returns the flight curves and the data (it returned the unformatted list).

Documentation

• Vignette: the installation code is shown but not run, the messages of flight_curve() are not printed, linewidth replaces the deprecated size of ggplot2 lines; author and date updated.
• LICENSE year 2022-2026.

Potential breaking changes

• Code using flight_curve_multi(), fit_bam_multi() or plot_flight_curve_multi() (exported in 1.3.0) needs rbms 1.3.0, or flight_curve().
• flight_curve() now stops with a clear message when a required column of ts_season_count is missing.

rbms 1.3.0

New features

• collated_index() gains GlmFit = c("glm", "newton", "ipf"). With "newton" or "ipf", the same weighted Poisson GLM (site and year effects) is fitted without building the model matrix and gives the same collated index as glm(), much faster with many sites (e.g. 0.2 s instead of 3 minutes with 2000 sites and 20 years). With "newton", the year effects and their standard error are returned (year_effect).
• collated_index() gains site_length (index per 1 km of transect) and index_sites = c("all", "observed") (index averaged over all monitored sites or over the sites where the species was observed).
• New collated_index_mixed(): the collated index from a mixed model with a random site effect (package glmmTMB, Poisson or negative binomial), with an optional habitat (stratum) effect and habitat weights (by sites, equal or user defined, e.g. area), site_length as offset and index_sites. It links years that are not linked through common sites, for randomised designs such as rotating panels.
• boot_sample() gains s_sp to draw the bootstrap samples per species, stratified by the occurrence of zeros at the sites and by group of linked years, with habitat as additional stratum. A sample with fewer than min_year years or where a group of linked years is split or lost is drawn again, up to max_trial times; failed samples are returned as NA rows. Without s_sp, the samples are drawn as before.
• impute_count() gains SindexMethod = c("ratio", "glm"). The default ratio is unchanged; with "glm", the counts of the time steps not visited are predicted from a Poisson GLM with site and year effects and the flight curve as offset (GlmFit = c("newton", "ipf")), which avoids extreme site indices at low coverage. The year effects are returned as attribute glm_year_effect.

Changes

• collated_index(): when the years form groups not linked through sites with observations, each group is fitted separately (SplitGroup = TRUE, default) and identified in the new column GROUP; fitting all years together gave arbitrary levels between groups. SplitGroup = FALSE reproduces the previous fit.
• Site-years with NA or infinite site index are treated as missing: impute_count() returns NA instead of Inf or NaN when the visits cover no part of the flight curve (TOTAL_NM = 0), and site_index(), collated_index(), collated_index_mixed() and boot_sample() remove them (previously a site with NA in all years set the collated index to 0 in all years).
• A bootstrap sample that failed in boot_sample() returns NA in collated_index(); boot_ind drawn for another species is an error.
• collated_index_old() is deprecated and returns an error, use collated_index().
• plot() and points() methods for pheno_curve objects: the first argument is now x, as in the generics.
• speedglm is no longer needed; glmmTMB and ggplot2 are suggested.

Potential breaking changes

With the default arguments, the results of the standard workflow (flight curve, imputed counts, site indices, collated index and bootstrap) are identical to version 1.2.1. A pipeline is only affected in the following, unlikely cases:

• collated_index() returns an additional column GROUP in col_index and an additional element year_effect; code that selects columns by position or binds outputs with rbind() may need adjusting (selection by name and data.table::rbindlist() are not affected).
• Data where the years form groups not linked through common sites: the collated index now differs, each group being fitted separately (SplitGroup = FALSE gives the previous result, whose levels between groups were arbitrary).
• Data with NA or infinite site indices: these site-years are now removed; previously an NA site set the collated index to 0 in all years, and site_index(MinFC = NULL) kept infinite values.
• speedglm is no longer installed with rbms; scripts calling speedglm() directly need it installed.
• collated_index_old() now returns an error (it did not run in previous versions).

Documentation

• Vignette: per-species stratified bootstrap, linked years and groups of years, when separate groups can be compared, the collated index as a density (per km, sites used) and references with DOIs.