Skip to content

Latest commit

 

History

History
142 lines (99 loc) · 9.78 KB

NEWS.md

File metadata and controls

142 lines (99 loc) · 9.78 KB
Raw

dymiumCore 0.1.9.9001

New features

  • ModelMultinomialLogit and ModelBinaryChoice now have S3 predict and summary methods. Note that ModelMultinomialLogit requires newdata to be in the same format that is required by mlogit.

Changes

  • Move matchingR to Suggests.
  • Replace lest::case_when() with data.table::case_when() (#57).
  • Add a name argument to Model's constructor function and expose it as an R6 active field.
  • World$add() can now be used to add a named Model without providing the name argument.
  • World$add() gained a replace argument with TRUE as its default value.

Bug fixes

dymiumCore 0.1.9

  • introduced ModelCustom a model class that let users specify its parameters and predict function. We also add ModelMultinomialLogit, ModelBinaryChoice and ModelLinear which are implementations of ModelCustom.
  • add makeModel to create a light weight Model object that can be used in the transition functions and classes. This basically creates an appropriate Model class from the given model object.
  • add helper functions which_min_n and which_max_n.
  • renamed unnest_datatable to unnest_dt which can take names of the list columns as a character vector.

dymiumCore 0.1.8

  • Added sim() for compiling and executing a microsimulation pipeline.
  • Added targets as an active field.
  • Fixed get_transition didn't use the preprocessing function when the model argument is of Model class.
  • get_transition now removes the dot prefix of variables in entity data. Those variables are considered as derived variables which will not be checked against when new entity data are geting added.
  • Added dsample for sampling from a discrete distribution, an extension of base:sample.
  • Added remove_entity for removing entities from an [Entity] object.
  • Entity$get_ids can get the ids of removed entities using newly added the include_removed argument.
  • Entity gains a primary_id field.
  • Entity$add did not check for potential duplicate ids in the removed data.
  • World$set_time returns self. The $start_iter method may be removed in the near future as it is not quite intuitive.
  • Fixed the unexpected behaviour when a non Target object is passed into target arg of Transition. Prior to this fix, the object that got passed into the target argument was used to create a Target object. Hence, this created an expected behaviour, when the Target's get method was called it returned a scaled down value of the target. This may not be the desire default effect by users.

dymiumCore 0.1.7

  • transition() gains values field, it expects a named vector where its values will be used to replace the outcomes of the model and update the variable in attr. See the example section to learnsmore.
  • transition() now supports a named list model, e.g: list(yes = 0.1, no = 0.9).
  • Introduced add_log(), get_log(), mutate_entity(), and add_entity(). These functions are supposed to make constructing a microsimulation pipeline simpler. See their documentation for details.
  • Made $colnames() into an active field for DataBackendDataFrame.
  • Added omit_derived_varnames() for internal use.
  • Individual now has a hid_col active field which returns its household id column name.
  • Entity$add() now gives newdata new unique ids by default.
  • Deprecated assign_reference(), Container$get() should be used instead.

Minor changes

  • Generic$log() now uses NA_character_ as the default in its tag argument.

dymiumCore 0.1.6

NEW FEATURES

  • Added transition(), this is a function for simulating state transition inside a microsimulation pipeline. It also comes with get_transition() which returns state transition outcomes. See its documentation, ?dymiumCore::transition, for more detail.
  • [Model] gains preprocessing_fn field, this is for storing a preprocessing function which will be used to evaluate the entity data in [Transition] prior to simulating the transition. A situation where this is useful could be when you want to limit the use of a [Model] object to the specific group of agents (e.g: age between x and y) that was used to estimate the model.
  • [Trans] supports [Model]'s preprocessing_fn().
  • Minor fix to error msg in [TransitionClassification].
  • [add_models] gains a as_r6model for returning models as [Model] objects.
  • Classification models fitted using the mlr package is now supported by transition.
  • Added simulate_choice, a function for simulating choice selection from the predicted probability obtained from a model object, currently only supports most of the classification models fitted with caret and mlr, and glm of the binomial family.

INTERNAL CHANGES

  • Renamed Transition to Trans to make way for transtion().

dymiumCore 0.1.5

NEW FEATURES

  1. Added Population$household_type method for classifying household types of household agents. There are four types: 'non_family_hh', 'couple_hh', 'couple_hh_with_children' and 'lone_parent_hh'. See its documentation for detail.
  2. Added a model active field to Model, this returns the original model object stored inside the object.
  3. Model gained a S3 summary method.

FIXES

  1. Household$update_hhsize() now returns NULL invinsibly instead of its attribute data.
  2. World now accepts Model objects in add.

dymiumCore 0.1.4

NEW FEATURES

  1. Added add(.data, ...) method to Entity. This allows new entities to be added to the attribute database of existing entities. Note that, add_data(.data) is for adding new databases and add(.data, ...) is for adding new records to the attribute database (Entity$database$attrs) which are not the same.
  2. Added check_subset2 which basically the same check as checkmate::check_subset but it returns a short error message. The error message it returns only include those missing elements in x from choices. See checkmate::check_subset for more details.
  3. DataBackendDataTable again on option to use key(s) for row indexing.
  4. Added Entity$get_data2(ids) which uses key(s) for subsetting of ids. This suppose to be a faster implementation of get_data() which is likely to supersede the original implementation in the next version.
  5. add_population(ind_data, hh_data) now also assigns new ids to all the records of new entities of ind_data and hh_data to make sure no duplications of ids exist.
  6. DataBackend has new active fields which are data amd removed_data these functions return a copy of the data and not a reference to the data (only applicable in DataBackendDataTable and DataBackendSpatialFeature).
  7. extract_data returns all the data objects in the DataBackend objects that each Entity possess as a named list of data.table. This is useful for saving simulation data for furthur analysis. It works on World too!
  8. World saves session info on its creation instead of just the R version it was created on.
  9. Added dymium.simulation_scale global option that can be set with World$set_scale(x) and access with World$scale or getOption("dymium.simulation_scale"). This simulation scale will be used by all Target objects created when they are called by their get method.

BUG FIXES

  1. Ignore checking of attributes in all data.table::all.equal(...) calls.
  2. Fixed Target's constructor method which failed to convert data.frame to data.table when storing the target data.

DEPRECATIONS

  1. Removed Entity$initialise_data(), the attribute data of Entity must be provided in its constructor method. This change affected many of the testthat tests.
  2. Removed household_formation() to encourage more explicit approaches (e.g. use Population$leave_household() and Population$join_household()).

dymiumCore 0.1.3

NEW FEATURES

  1. Added a plot_relationship method to Household. This uses visNetwork for plotting (added to Suggests). See #48 for its implementation detail.
  2. inspect now has a verbose option.
  3. Transition no longer removes the NA reponses when target is used.
  4. Added a replace method to World which basically remove and add in one call.
  5. Moved $subset_ids() from Agent to Entity.
  6. download_module() and set_active_scenario() now have a .basedir argument which sets the base directory where their files will be created at. By default this is the root folder of the currently active R project (if you are using RStudio) which is determined by here::here().
  7. Renamed use_scenario to create_scenario and active_scenario to get_active_scenario.
  8. TransitionClassification's target argument now accepts a dynamic target, see issue [#52] #52.
  9. Added a Target R6 class which acts as a wrapper for different types of target and make them work consistently in the Transition classes.

BUG FIXES

  • Fixed World's remove method in commit c0df66d.
  • Removed the preprocess method in Transition. Whether the mutate step or the filter step should be run first is now determine by the mutate_fisrt public field which has FALSE as default.
  • create_scenario did not pass .basedir to set_active_scenario so when creating a new scenario that isn't within a RStudio project will raise an incorrect directory error.

DEPRECATIONS

  • Unexported Firm as it has not been fully implemented and tested yet.
  • Removed deprecated Population's methods: get_sum_hhsize and count_all.

dymiumCore 0.1.2

  • TransitionClassification now treats probs as probability weights and not exact probability values that should add up to 1.

dymiumCore 0.1.1

  • Removed the target argument from TransitionRegression as alignment is only applicable for classficaition model.

dymiumCore 0.1.0

This version contains all the basic building blocks for dynamic microsimulation.