deepSTRAPP 1.1.0
- Handle uncertainty in trait estimates. See the
‘uncertainty_strategy’ argument in run_deepSTRAPP_*() functions.
- Add functions to load results from external BAMM analyses. See the
dedicated vignette/tutorial “import_external_analyses”.
Accounting
for uncertainty in ancestral trait estimates
- STRAPP tests can now be run across a posterior sample of trait
histories rather than a single reconstruction. The
uncertainty_strategy argument of
run_deepSTRAPP_for_focal_time(),
run_deepSTRAPP_over_time() and
compute_STRAPP_test_for_focal_time() selects how trait and
rate uncertainty are combined: "rates_only" (BAMM posterior
samples with a single trait reconstruction), "paired" (each
stochastic map paired with one BAMM sample), or "full" (all
stochastic maps crossed with all BAMM samples).
- Stochastic maps can be supplied directly through the new
contMaps and simmaps arguments, or simulated
from posterior probabilities with nb_simulations.
trait_maps_vs_BAMM_samples_list records, and lets you
impose, which stochastic map is paired with which BAMM sample, so that
the same pairing is used at every time step of a trajectory.- See the new vignette
handle_uncertainty.
Importing results from
external analyses
- New functions to bring analyses run outside deepSTRAPP into the
workflow:
build_BAMM_object(),
subset_BAMM_object() and prune_BAMM_object()
for BAMM output; convert_BSM_to_simmap() and
convert_BSMs_to_simmaps() for BioGeoBEARS biogeographic
stochastic maps; convert_contsimmap_to_contMaps() for
contsimmap output; and
convert_simmaps_to_densityMaps() to summarise stochastic
maps as posterior densities.
aggregate_contMaps() summarises a list of continuous
stochastic maps into a single mean or median contMap.- See the new vignette
import_external_analyses.
Consistency
of the statistical method across time steps
run_deepSTRAPP_over_time() now selects the statistical
method once, before any test is run, from all states/ranges described in
the complete trait mapping, and applies it at every time step. The
method previously depended on the states/ranges still present at each
time step, so a trajectory could mix Kruskal-Wallis and Mann-Whitney U
p-values on a single curve when a state/range was absent from the deeper
time steps. This is not the case anymore, and all p-values across
time-steps are prodcued by the same type of test.- The states/ranges actually observed are now reported:
$states_observed and $nb_states_observed per
time step, and $states_observed_overall,
$states_observed_per_time_steps and
$nb_states_observed_per_time_steps in the output of
run_deepSTRAPP_over_time().
Other additions
- New
extract_all_trait_values_for_focal_time() to
extract trait data from every stochastic map at a given time, and
extract_trait_data_melted_df_for_focal_time() to return it
in long dataframe format.
- New
cut_contMaps_for_focal_time(),
cut_simmap_for_focal_time() and
cut_simmaps_for_focal_time() to cut lists of stochastic
maps at a focal time.
run_deepSTRAPP_for_focal_time() and
run_deepSTRAPP_over_time() gain
return_updated_Maps to return the mappings cut at each
focal time, and run_deepSTRAPP_for_focal_time() gains
extract_trait_data_melted_df to return the underlying trait
data in long dataframe format.
deepSTRAPP 1.0.0