Changes in version 0.4.0                        

Saqrlab is now simulation-only: network estimation code (build_network,
bootstrap_network, permutation_test, temporal_network, mlvar, GIMME,
MCML, higher-order networks, etc.) has moved to the sibling package
Nestimate. This release focuses the package on synthetic-data generation
with known ground truth.

New Features

Unified saqr_sim interface

  - New saqr_sim S3 class wraps every numerical simulator's output with
    a consistent $data / $params / $type / $seed interface, plus [,
    as.data.frame(), head(), tail(), dim(), names(), str(), print(), and
    summary() methods. $data / $params access is backward-compatible.
  - simulate() — single dispatcher that routes a type string to the
    matching explicit-parameter simulator.
  - list_simulators() — catalogs the available simulation types as a
    data.frame.
  - validate_recovery() — compares a method's point estimates to the
    known true params, returning a tidy per-parameter recovery table
    (absolute/relative error and a within-tolerance flag).

Explicit-parameter simulators

  - simulate_ttest(), simulate_anova(), simulate_correlation(),
    simulate_clusters(), simulate_prediction() — generate data from
    user-specified means, SDs, correlation/covariance matrices, cluster
    centers, and regression coefficients; ground-truth effect sizes
    returned in params.
  - simulate_longitudinal() — VAR(1) multilevel panel data for mlVAR/ESM
    with explicit temporal, contemporaneous, and between-person
    matrices.

New simulation modules

  - inject_missingness() — adds MCAR, MAR, or MNAR missingness to any
    simulated data frame at a controlled rate.
  - simulate_mlm() and simulate_growth() — multilevel and latent-growth
    / longitudinal data from explicit fixed effects and variance
    components.
  - simulate_irt() — item response theory data (1PL/2PL/3PL) from
    explicit item parameters and latent abilities.
  - simulate_survival() — time-to-event data with explicit hazard,
    covariate effects, and censoring.
  - simulate_hmm() — hidden Markov model sequences from explicit
    emission and transition probabilities.

Scenario presets

  - list_scenarios(), get_scenario(), and run_scenario() — named
    simulation recipes for common designs.
  - tidy_simulation_results() — flattens simulation results to a
    data.frame.
  - export_simulation() — writes results to disk.

Testing

  - Added a cross-package fixture-contract test that guards the
    simulation output contract relied on by downstream packages.

Bug Fixes

  - Fixed simulate_group_tna_networks() (group TNA network generation).
  - Fixed compare_networks() / network-comparison summaries.
  - Fixed grid-simulation result summaries.

                        Changes in version 0.1.0                        

Initial release of Saqrlab - Simulation and Analysis Tools for Temporal
Network Analysis.

New Features

Data Simulation

  - simulate_matrix() - Generate simple transition matrices with
    learning state names
  - simulate_htna(), simulate_mlna(), simulate_mtna() - Multi-type
    matrices for hierarchical/multilevel network analysis
  - simulate_sequences() - Generate Markov chain sequences with optional
    learning states
  - simulate_sequences_advanced() - Sequences with stability modes for
    realistic patterns
  - simulate_long_data() - Hierarchical group data with actors, groups,
    and courses
  - simulate_onehot_data() - One-hot encoded format for sequence data
  - simulate_edge_list() - Social network edge lists simulation

Network Generation

  - simulate_tna_datasets() - Complete TNA datasets with sequences and
    parameters
  - simulate_tna_networks() - Fitted TNA models with random parameters
  - simulate_group_tna_networks() - Group TNA models for multi-group
    analysis
  - simulate_tna_matrix() - HTNA/MLNA matrices with node types
  - generate_probabilities() - Random transition matrices and initial
    probabilities

Model Fitting

  - fit_network_model() - Fit TNA, fTNA, cTNA, and aTNA models

Network Comparison

  - compare_networks() - Compare two networks using correlation, RMSE,
    MAE, and more
  - compare_centralities() - Compare centrality profiles between
    networks
  - compare_edge_recovery() - Edge recovery metrics (precision, recall,
    F1)

Batch Processing

  - batch_fit_models() - Fit models to multiple datasets in parallel
  - batch_apply() - Apply any function to a list of objects

Bootstrap & Simulation Studies

  - run_bootstrap_simulation() - Bootstrap analysis for stability
    testing
  - run_grid_simulation() - Parameter grid search simulations
  - run_network_simulation() - Comprehensive model comparison studies
  - run_bootstrap_iteration() - Evaluate single bootstrap runs
  - summarize_grid_results() - Analyze grid simulation output

Learning States

  - get_learning_states() - Get learning verbs by category
  - list_learning_categories() - Show available categories with counts
  - select_states() - Intelligent state selection based on network size
  - LEARNING_STATES - Full dataset of 180+ learning verbs in 8
    categories
  - GLOBAL_NAMES - 300 diverse names for simulation
  - get_global_names() - Retrieve names from the global names dataset

Utilities

  - generate_param_grid() - Create parameter combination grids
  - validate_sim_params() - Validate and set defaults for simulation
    parameters
  - summarize_simulation() - Summary statistics for simulation results
  - summarize_networks() - Network-level summaries for model lists

Improvements

  - Standardized parameter names across all functions
  - use_learning_states = TRUE as default for realistic educational
    simulations
  - Parallel processing support via future and future.apply
  - Progress reporting with progressr

Dependencies

  - Core: tna, seqHMM
  - Data manipulation: dplyr, tidyr
  - Parallel processing: parallel, future, future.apply
  - Utilities: progressr, lhs