NEWS.md

In TilburgNetworkGroup/remstats: Computes Statistics for Relational Event History Data

remstats 3.2.2

Date: May 14, 2024

New features

• [Issue [#56]] - The ability to generate exploratory plot() and boxplot() for remstats objects. These functions were designed to work with large remstats objects by taking a subset of timepoints, dyads or actors (depending on the type of plot). The user can determine the subset to be taken. Setting the subset to the whole sequence is not recommended for large remstats objects.
• Added warning when consider_type is specified but event types are not in the DV.
• Added tests for plotting functions.
• Increased test coverage.

Fixed

• Resolved conversion errors occurring when using single-letter logical values (i.e., 'T' for TRUE or 'F' for FALSE).
• Resolved error with switch on boolean type (in tomstats.cpp)

remstats 3.2.1

Date: November 29, 2023

Fixed

• Fixed [Issue [#78]]: The attributes of the remstats object are now correctly set after executing bind_remstats().
• Resolved a bug where remstats failed to find the edgelist when reh is ordinal.
• Resolved a bug where aomstats failed when the computation method was pe, the memory method was decay and the first event at the start index was not the first event at that time.
• Resolved Rcpp string warning.

remstats 3.2.0

Date: November 16, 2023 Compatible with remify version 3.2.0.

New features

• [Issue #67] - Introducing the ability to manage simultaneous events. When events occur on the same time, the user can specify the method for managing simultaneous events. The default method is "pt" (per timepoint), where statistics are computed once for each unique timepoint in the edgelist. Alternatively, you can choose "pe" (per event), where statistics are computed once for each unique event observed in the edgelist.
• [Issue [#57]] - Character / factor variables are allowed for the same statistic.
• The subset and method object to the attributes of tomstats and aomstats objects.

Changed

• The meaning of the start and stop values changes slightly when method is pt. They now refer to the indices of the first unique timepoint and last unique timepoint for which the user wishes to calculate statistics. The meaning of start and stop remains unchanged when method is pe.
• [Issue #72] - When weighting the past events, counting now starts from the time of the preceding event instead of the current event. This change ensures the constant hazard assumption is maintained.
• The default value for the consider_type argument is now TRUE for improved logical consistency.
• Added information about the methods to the printed remstats object.

Fixed

• Bug in formula attributes of aomstats object.

Removed

• The ccp effect has been removed.

remstats 3.1.8

Date: October 25, 2023

• Works with remify >= 3.1.0.

Fixed

• Resolve the bug in bind_remstats() that occurs when combining two actor-oriented remstats objects, one of which has receiver effects while the other does not.

remstats 3.1.7

Date: October 9, 2023

• Works with remify >= 3.1.0.

New features

• Flexible data format: Introducing support for long-format data in the tie() statistic.
• Temporal variability: Introducing support for time-varying covariates in the tie() statistic.
• Alternative statistic: Presenting the dyad() statistic as an alternative to the tie() statistic.

Major changes

• Updated function arguments: To align with the introduction of the long data format (attr_dyads) for the tie() statistic, the previously used attr_data object has been renamed to attr_actors. This change ensures consistency and clarity in our terminology.

remstats 3.1.6

Date: October 6, 2023

• Works with remify >= 3.1.0.

New features

• Collection of participation shifts are available for the actor-oriented model.

remstats 3.1.5

Date: September 5, 2023

• Works with remify >= 3.1.0.

Bug fixes

• bind_remstats() now works for the actor-oriented model

remstats 3.1.4

Date: August 28, 2023

• Works with remify >= 3.1.0.

Bug fixes

• remstats() now works for ordinal event sequences.

remstats 3.1.3

Date: July 13, 2023

• Works with remify >= 3.1.0.

New features

• display_progress argument for the tie-oriented model.

remstats 3.1.2

Date: July 6, 2023

• Works with remify >= 3.1.0.

New features

• unique argument in triadic statistics

Major changes

• Naming of the interaction effects from var1.x.var2 to var1:var2.
• spUnique is deprecated. Instead sp(unique = TRUE) can be used. This is done to be consistent with the new unique argument in the directed triadic statistics.

Minor changes

• Switch from testthat to tinytest.

remstats 3.1.1

Date: July 1, 2023

• Works with remify >= 3.1.0.

New features

• bind_remstats()
• Compute statistics for only the "active" dyads in the risk set (in combination with remify, see remify::remify riskset argument)

Major changes

• The attributes argument that exists in many of the function is renamed to attr_data to avoid naming conflicts.
• Change of the scaling = "as.is" option to scaling = "none".

Minor changes

• More efficient computation triadic statistics

remstats 3.1.0

Date: June 1, 2023

IMPORTANT: remstats 3.1.0

• Works with remify >= 3.0.0. Make sure remify is updated. If earlier versions of remify are in use, remstats 3.1.0 will break down.

New features

• print(), and summary() methods.

Major changes

• edgelist argument is renamed to reh.
• reh must be an object of class remify.
• id column in attributes object is renamed to name.
• aomstats output is reduced to the list of statistics (sender_stats and receiver_stats)
• tomstats output is reduced to the array of statistics. The risk set is now an attribute. The adjmat is only an attribute if requested with "get_adjmat" (this is changed to save memory).

Minor changes

• Integrated with remify v3.0.0
• Updated tests.

remstats 3.0.3

Date: March 16, 2023

New features

• Interval memory.
• Obtain a list with available effects with the functions tie_effects() and actor_effects().
• Show progress with display_progress argument in aomstats() (will be added to remstats() and tomstats() later.)

Minor changes

• Fixed bug in exogenous stats with time-varying attributes in the tie-oriented model.
• Fixed bug in scaling the exogenous stats in the receiver step of the actor-oriented model.
• Added explanatory warnings and errors.
• Fixed partial match warning (issue #35).
• Fixed environment issue attributes object.
• Renamed "Brandes" memory to "decay".
• Updated computation degree, inertia, reciprocity and triadic statistics in actor-oriented model: without adjacency matrix (issue #39).

remstats 3.0.2

Date: February 8, 2023

New features

• psABAB() effect
• psABAY() effect adapted for undirected events
• degreeDiff() effect

remstats 3.0.1

Date: January 5, 2023

New features

• userStat() effect

remstats 3.0.0

Date: December 22, 2021

IMPORTANT: remstats 3.0.0

• Works with remify >= 2.0.0. Make sure remify is updated. If earlier versions of remify are in use, remstats 3.0.0 will break down.

Major changes

• Integrated new version (2.0.0) of remify.

Minor changes

• Fixed bug in tie() (see issue #29).
• Added warning messages when redundant actors are present in input data for the exogenous stats (see issue #30).

remstats 2.0.4

Date: December 21, 2021

New features

• Enabled ARMA_64BIT_WORD

remstats 2.0.3

Date: December 16, 2021

New features

• Added degreeMin and degreeMax statistics for undirected events.
• Added ccp statistic for undirected, dyadic events.
• Option to output only statistics (since outputting all objects may take a lot of time).

Major changes

• Updated computation ``Brandes memory'' in adjmat to include normalization factor.

remstats 2.0.2

New features

• Added verbose argument that, when set to TRUE, outputs a progress update on the statistics computation.
• Added vector with types names to output object of tomstats().

Major changes

• Changed the specification of the variable in the event() effect.

Minor changes

• Updated computation procedure for triadic and rrank statistics in the tie-oriented model (tomstats) for greater efficiency and less computation time.
• Changed the way interaction dimnames of the statistics are defined so its more informative.
• Updated documentation.

remstats 2.0.1

New features

• Added the remstats() function, which is a wrapper for aomstats() and tomstats().
• Added vector with actor names to output object.

Major changes

• Changed aomstats() function arguments that refer to the requested effects: sender_effects (was rateEffects) and receiver_effects (was choiceEffects).
• Changed names of the aomstats() statistics List output object to sender_stats (was rate) and receiver_stats (was choice).
• Fixed bug in computation "spUnique" effect (was affected by event weights).
• Added variableName tie to dimnames statistics object.
• Fixed bug in computation "event" effect in combination with windowed memory (covariates object was not sliced).
• Fixed bug in computation "tie" effect (wrong ordering).

Minor changes

• Updated package description.
• Updated functions documentation.

remstats 2.0.0

• Added a NEWS.md file to track changes to the package.

New features

• The function aomstats() is added to compute statistics for the actor-oriented model. Effects for the rate-step and the choice-step of this model have to be specified separately, see the function's documentation or the README.md.
• The recency statistics have been extended and include the following:
  • recencyContinue: refers to the time that has past since dyad (i,j) last interacted.
  • recencyReceiveReceiver: refers to the time that has past since receiver j last received an event.
  • recencyReceiveSender: refers to the time that has past since sender i last received an event.
  • recencySendReceiver: refers to the time that has past since receiver j last sent an event.
  • recencySendSender: refers to the time that has past since sender i last sent an event.
• The option Brandes is added to the memory functionality. This refers to the exponential decay of the weight of past events, depending on the time that has past since the event occurred. The memoryValue argument refers to the halftime parameter (see Brandes et al., 2009).

Major changes

• The name of the function remstats(), which was the main function in the previous version, is changed to tomstats() (because it computes statistics for the tie-oriented model, as opposed to the function aomstats(), which computes statistics for the actor-oriented model).
• memory and memoryValue are added as arguments of tomstats() and aomstats() and removed from the separate statistic functions. This is because based on these memory settings, an adjacency matrix is computed. Based on this adjacency matrix, a lot of the endogenous effects are computed. While this increases the efficiency of the computation, memory effects can no longer be specified separately per endogenous variable.
• event weights have to be specified as a separate weight column in the edgelist that is supplied to tomstats() or aomstats() instead of to a separate argument in the separate statistic functions. This is because based on these weights, an adjacency matrix is computed that is used for the computation of the endogenous effects. (see previous point).
• Effects that are requested should be specified in the effects argument (tomstats()) or rateEffects and choiceEffects argument (aomstats()). Previously, this argument was called formula.
• The actor and event types in the edgelist and riskset output of tomstats() and aomstats() are transformed back to their original ID's, instead of to the ID's used by remify::reh() and internally, so that the user can more easily assess the computed statistics.
• An intercept is specified in the same manner as in lm(). In the tie-oriented model and rate step of the actor-oriented model, an intercept term is assumed by default (unless the ordinal argument is set to TRUE). Alternatively, it can be explicitly specified by adding 1 to the terms of the effects formula or explicitly removed by adding -1 to the terms. The baseline() specification is removed.
• The exogenous statistic equate() has been removed, something similar can be achieved with tie().
• Fixed effects for the event type can now be specified with FEtype() (could previously be done with baseline(with_type = TRUE)).
• The argument with_type in the endogenous statistic functions is renamed to consider_type.
• Previously, two or more variables could be specified in one exogenous effect formula. This functionality is removed, only one variable per exogenous effect formula can be specified.

Minor changes

• An attributes object (previously named covariates) can now be supplied to the main functions tomstats() or aomstats() directly, but can also still be specified in the separate functions for exogenous statistics.
• To increase efficiency, computation of some of the statistics is based on an adjacency matrix that is either first computed internally or can be supplied by the user. The user won't notice much from this, except that the adjacency matrix (if computed) is also outputted and can be inputted again to decrease computation time if an extra statistic is requested.

TilburgNetworkGroup/remstats documentation built on Jan. 31, 2025, 12:38 p.m.