NEWS.md
This is a maintenance to release to ensure compatibility with the upcoming release of purrr. There aren’t any major changes. A couple features have been added, and some bugs have been fixed.
as.matrix() method for ipmr_ipm and ipmr_matrix. This strips the ipmr_matrix class away, and returns a matrix/array object. When called on an ipmr_ipm, it returns a list of sub-kernels that are just standard matrix/array objects. Thanks for @Aariq for requesting this (#63).The title in conv_plot is now correctly updated when there are multiple parameter sets used to construct an IPM. Previously, the title was always something like lamdba_1, regardless of which level of parameter set was being plotted. Thanks to @SanneE1 for pointing this out (#64).
Clarifies some error messages.
Removes automatic faceting in plot.*, as that usually failed when too many kernels were present. Instead, just use the nrow and ncol arguments to the function now.
Contains a number of bug fixes and some new functionality. The latter are mostly related to PADRINO models, and shouldn’t have too much of an effect on existing user-specified IPMs.
Adds return_sub_kernels argument to make_ipm() for *_stoch_param and all density dependent methods. This is due to the large RAM footprint that these objects can occupy, particularly for long running models. The default is FALSE, which will save considerable memory space. Can be set to TRUE for downstream analyses that require sub-kernels/iteration kernels.
Prettier printing method for PADRINO objects/lists of user-specified models.
Prettier warnings for left/right_ev()
Depending on your view, this may be a bug fix: updates the log argument in lambda() so that it only changes the scale, NOT the calculation method. The prior behavior was documented, but unlikely to be intuitive, and so caused some confusion. Thanks to @aariq for pointing this out.
For stochastic IPMs, is_conv_to_asymptotic() and conv_plot() now check for convergence in stochastic lambda. That is, they use a cumulative mean log(lambda) over iterations after discarding a burn-in. Thanks to @aariq for implementing this in #45.
Adds experimental function make_ipm_report(). This function converts proto_ipm objects into Rmarkdown documents that, when rendered, contain the equations and parameters used to implement the model in Latex. See the function documentation for more details, and report bugs/notation quirks/preferences in the issue tracker.
Fixes bug in normalization of left/right eigenvectors in simple density independent stochastic models. (thanks to @aariq)
Fixes bug where "drop_levels" was not recognized in some parameter set indexed models.
Fixes bug where parameter set levels were recycled in some cases.
Fixes bug in lambda() where log = TRUE had no effect when the IPM was stochastic and type_lambda was 'last' or 'all'.
Switched from all.equal to absolute tolerance. @aariq in #52.
Fixes bug where the value of lambda() was named for determinstic models and unnamed for stochastic models. @aariq in #57.
Fixes bug in make_ipm() where a user-specified sequence wasn’t getting used correctly for certain model classes. @aariq in #58.
Warnings about NAs in data_list are no longer raised for model objects.
Contains small tweaks, bug fixes, and new feature additions. There shouldn’t be any breaking changes to the API.
Adds log argument to lambda so users can choose which scale to return. The default is TRUE for stochastic models and FALSE for deterministic models.
Allows named expressions in define_env_state() such that the names of the expressions may be used in define_kernel(). Before, the name of the expression didn’t matter, only the names of the outputted list. Now, this will also work as well.
define_env_state(proto_ipm,
temp = rnorm(1, 20, 3),
precip = rgamma(1, 400, 2),
data_list = list())More unit tests for parameter resampled models.
Removed innocuous warning messages. Added warnings for NA values in a few define_* functions and errors when they’re produced in sub-kernels.
Added the ipmr_ipm class so that most functions can tell the difference between a PADRINO object and an ipmr object.
Changes argument tol to tolerance in is_conv_to_asymptotic() for consistency with other function/argument names.
corrects a bug where functions in the define_env_state data_list argument weren’t recognized.
corrects bug in left_ev() and right_ev() where parameter set indices were ignored for deterministic models .
Contains a some tweaks and bug fixes, and a few new features:
Implements right_ev() and left_ev() methods for stochastic models.
Adds a new function, conv_plot(), to graphically check for convergence to asymptotic dynamics in deterministic models.
Adds a new function, discretize_pop_vector(), to compute the empirical density function for a population trait distribution given a set of observed trait values.
Adds print methods for density dependent models.
Adds log argument for lambda.
Corrects bug where tol argument was ignored in is_conv_to_asymptotic().
Gives output from lambda() names. Before, outputs from deterministic models with many parameter sets became hard to follow.
Contains a some tweaks and bug fixes. There is one major API change that renames parameters in define_kernel.
Renames function arguments hier_effs -> par_sets, levels_ages/levels_hier_effs -> age_indices/par_set_indices. The idea was to shift from thinking about these IPMs as resulting from multilevel/hierarchical regresssion models to IPMs constructed from parameter sets (which can be derived from any number of other methods).
Corrects some bugs that caused vital_rate_funs() to break for stoch_param and density dependent models.
Updates the age X size model interface so that max_age kernels can be specified separately if they have different functional forms from their non-max_age versions.
make_iter_kernel can handle computations passed into mega_mat (e.g. mega_mat = c(P + F, 0, I, C)).
Makes plot.ipmr_matrix more flexible, which is now the recommended default plot method for ipm objects.
Changes to internal code that won’t affect user experience.