LoginSign Up
← Back to all sparks
M

modeldata

ANALYTICS
Velocity0.0

Curated data sets used for teaching and testing tidymodels workflows

github.com/tidymodels/modeldata ↗

The tidymodels example-data package grows one dataset at a time, on nobody's schedule

tidymodelsexample-datasimulationteaching-datar-package
◆Current state
modeldata exists to supply the datasets and simulation functions that tidymodels documentation, tests, and teaching material depend on. Releases arrive roughly once or twice a year and consist almost entirely of new data sets plus occasional simulation methods. The most recent work adds a Worley (1987) regression simulation and moves the package off the magrittr pipe onto base R's.
◆Where it's heading
Two lines run through the history: broadening coverage of task types — ordinal classification, multinomial, regression, QSAR-style chemistry data — and building out synthetic simulation so tutorials can demonstrate a method without shipping a real dataset for it. The simulation side has grown from a single regression generator into a family with logistic and multinomial variants and a keep_truth option that exposes the error-free outcome. Infrastructure changes appear only when the wider tidyverse moves, as the base-pipe transition shows.
◆Prediction
The pattern points to another simulation method or a dataset filling a task type the collection still lacks, rather than any change in what the package does.

◆Recent moves

  1. 11mo ago

    modeldata 1.5.1 fixes documentation and column-name typos

    A small release correcting typos in documentation and column names. Housekeeping between substantive data additions.

    View source ↗
  2. 1y ago

    modeldata 1.5.0 adds a Worley (1987) regression simulation

    Adds a Worley (1987) regression simulation method and moves the package from the magrittr pipe to the base R pipe. The simulation family keeps expanding, which is the half of this package that gets genuine design attention.

    View source ↗
  3. 2y ago

    modeldata 1.4.0 adds the cat_adoption data set

    Adds the cat_adoption data set. A single-purpose release of exactly the kind this package exists to make.

    View source ↗
  4. 2y ago

    modeldata 1.3.0 adds the deliveries data set

    Adds the deliveries data set, another regression example for documentation and tutorials.

    View source ↗
  5. 3y ago

    modeldata 1.2.0 adds eight data sets across regression and classification

    Eight new data sets land at once, spanning regression, classification, and ordinal classification, alongside a correction to the Hooker (2004) simulation equation. The largest single expansion in the visible history, and it deliberately covers task types the collection was thin on.

    View source ↗
  6. 3y ago

    modeldata 1.1.0 adds logistic and multinomial simulation plus keep_truth

    Supervised simulation functions gain a keep_truth argument exposing the error-free outcome as .truth, and sim_logistic() and sim_multinomial() extend the family past regression. This is where the simulation side stopped being a single generator.

    View source ↗