Skip to contents

Changelog

Source: NEWS.md

remstimate 2.3.11

  • minor fixes on implicit conversions

remstimate 2.3.10

  • minor fix on switch function for boolean type (remstimate.cpp).

remstimate 2.3.9

CRAN release: 2024-05-13

  • minor fix on titles of Schoenfeld’s residuals plots;
  • correction on learning rate parameter of “GDADAMAX” method.

remstimate 2.3.8

CRAN release: 2023-12-20

  • DESCRIPTION file, removed quotes from acronyms;

  • documentation (return values added where missing);

  • tests (removed test on ‘ncores’ argument);

  • vignette (switched to ‘html_document’, edited seed value and parameters for "HMC" method). . # remstimate 2.3.7 (initial CRAN release)

  • trace plot for posterior draws (only via "HMC" method) with highest posterior density intervals;

  • histograms of posterior draws (for both "BSIR" and "HMC" method) with highest posterior density intervals;

  • the content of summary.remstimate() can be saved by assigning it to a variable;

  • Watanabe-Akaike’s Information Criterion (WAIC) calculation available. New argument, logical value, WAIC, with default FALSE. If WAIC=TRUE, then an optional argument nsimWAIC (number of draws to estimate the WAIC) is set by default to 500 (the user can supply a different value);

  • updated parallelization of routines as to the computation of likelihood, gradient and hessian in the actor-oriented modeling framework;

  • updated parallelization of routines for computing model residuals and WAIC;

  • added new tests, increased coverage.

remstimate 2.3.6 - 2.3.1

  • several minor updates before initial CRAN release remstimate 2.3.7.

remstimate 2.3.0

  • Simultaneous events are supported (likelihood functions adapted).

remstimate 2.2.0:

  • major updates of Rcpp functions;
  • now using tinytest for testing (all tests are converted from testthat);
  • new examples to all functions in the package;
  • remove Rcpp functions (emp_dist_longest_batch(), for tuning L parameter in "HMC" method);
  • vignette created for actor-oriented modeling framework.

remstimate 2.1.0:

  • new remstimate logo;
  • update of functions’ description;
  • major update of remstimate() function:
    • removed input argument model;
    • initial controls of input argument;
    • output structure;
    • "HMC" for actor-oriented framework;
    • added initial values of "HMC" method (experimental);
  • major updates of Rcpp functions:
    • new experimental Rcpp function set_seed;
    • input argument edgelist changed with dyad;
    • new input arguments actor1 and actor2 (for actor-oriented routines);
    • removed old Rcpp functions: getDyadIndex(), getDyadComposition();
    • added dependency of exported C++ routines from remify, using now remify::getDyadIndex() and remify::getDyadComposition();
  • added new tests using testthat. Tests are divided in multiple .R scripts depending on what is tested;
  • update summary and print of actor-oriented framework in summary.remstimate;
  • predict.remstimate and plot.remstimate function are not yet available;
  • added dependencies: remify and remstats;
  • added data for tie-oriented and actor-oriented examples;
  • new README.md with badges.

remstimate 2.0.0:

  • This version is adapted to the latest changes coming from remify 2.0.0 and it can estimate a Tie-Oriented model as well as an Actor-Oriented model. Models can be estimated by means of different methods: "MLE", "GDADAMAX" (replacing the former "GD" and "GDADAM"), "BSIR" and "HMC". Methods like "BSIR" and "HMC" are ready-to-use but still under a continuous development in order to improve the user-experience;
  • removed experimental “fast method” to compute the likelihood;
  • added dependencies trust and parallel;
  • experimental Rcpp functions posteriorRank() and remDerivativesStandard_lambdas() (added but finally removed in the version 2.0.0);
  • added new Rcpp functions: getDyadIndex(), getDyadComposition().

remstimate 1.0.0:

  • 11/12/2020:
    • Methods working with the function remstimate() are: "MLE", "GD", "GDADAM", "BSIR", "HMC". However the output lacks of a structure attributes and methods;
  • 04/09/2020 :
    • messages becomes again an Rcpp file. This exstension appears to suit better the intent of the content/aim of error and warning messages;
    • remstimate.R contains the main function remstimate() which is aimed to run either a Frequentist or a Bayesian approach by using different optimization/methods. It also includes a switch to the “fast method” to compute the likelihood. The “fast method” is run if the actual improvement (percentage of improvement) is higher than a threshold set by the user (default is 0.5);
    • reh.cpp : contains getRisksetMatrix(), getRisksetCube(), convertInputREH(), getBinaryREH() and reh(). This last function is the one that preprocesses the input given by the user which consists in: edgelist, riskset and covariates. intereventTime variable and covariates input still need to be preprocessed via specific utility functions;
    • remstimate.cpp : contains remDerivatives() (which returns the value of loglikelihood, gradient, hessian at a specific parameter value), lpd() (log-pointwise density), utility functions for the fast method (cube2matrix(), getUniqueVectors(), computeTimes(), computeOccurrencies()) which is run with the function remDerivativesFast();
    • Since compute_stats() is not an exported function in remstats, getStats.R / compute_stats.cpp / compute_stats.h are temporary files so as to calculate statistics, run the estimation and compare estimates with relevent::rem(). getStats() is the alias of remstats::remstats() with some modifications at the stage of the preprocessing of the network;
  • 07/07/2020 :
    • reh() will be the only preprocessing function and it is coded in Rcpp (see reh.cpp file) whereas the R function and the reh.R file are removed;
    • utility functions called inside reh() are added inside the reh.cpp file, before the reh() function itself;
    • messages.cpp becomes a header file messages.h and the aim/content remains the same;
  • 03/07/2020 :
    • reh.h changed to reh.cpp and it contains utility functions used in reh.R within the R function reh(...) ;
    • messages.cpp will contain functions errorMessage(cond) and warningMessage(cond) that will return appropriate error/warning messages according to the cond argument;
  • 11/06/2020 :
    • Created reh.h were the utility functions to preprocess data will be developed;
    • remstimateBoost.h will contain the routines that speed up the computation of the loglikelihood and its first and second derivatives;
  • 20/04/2020 :
    • created repository with first commit;
    • package only contains three functions: remCpp(...) (uses optim to find the maximum likelihood estimates of REM), nllik(...) (returns the negative log-likelihood value for an observed event sequence, by specifying a vector of parameters and statistics), lpd(...) (calculates the same as nllik(...) but only for a specific time point and without taking the negative of the value).